xfs_aops.c

linux-2.6.15.6

out:
	unlock_page(page);
	return NULL;
}

STATIC int
xfs_map_unwritten(
	struct inode		*inode,
	struct page		*start_page,
	struct buffer_head	*head,
	struct buffer_head	*curr,
	unsigned long		p_offset,
	int			block_bits,
	xfs_iomap_t		*iomapp,
	struct writeback_control *wbc,
	int			startio,
	int			all_bh)
{
	struct buffer_head	*bh = curr;
	xfs_iomap_t		*tmp;
	xfs_ioend_t		*ioend;
	loff_t			offset;
	unsigned long		nblocks = 0;

	offset = start_page->index;
	offset <<= PAGE_CACHE_SHIFT;
	offset += p_offset;

	ioend = xfs_alloc_ioend(inode);

	/* First map forwards in the page consecutive buffers
	 * covering this unwritten extent
	 */
	do {
		if (!buffer_unwritten(bh))
			break;
		tmp = xfs_offset_to_map(start_page, iomapp, p_offset);
		if (!tmp)
			break;
		xfs_map_at_offset(start_page, bh, p_offset, block_bits, iomapp);
		set_buffer_unwritten_io(bh);
		bh->b_private = ioend;
		p_offset += bh->b_size;
		nblocks++;
	} while ((bh = bh->b_this_page) != head);

	atomic_add(nblocks, &ioend->io_remaining);

	/* If we reached the end of the page, map forwards in any
	 * following pages which are also covered by this extent.
	 */
	if (bh == head) {
		struct address_space	*mapping = inode->i_mapping;
		pgoff_t			tindex, tloff, tlast;
		unsigned long		bs;
		unsigned int		pg_offset, bbits = inode->i_blkbits;
		struct page		*page;

		tlast = i_size_read(inode) >> PAGE_CACHE_SHIFT;
		tloff = (iomapp->iomap_offset + iomapp->iomap_bsize) >> PAGE_CACHE_SHIFT;
		tloff = min(tlast, tloff);
		for (tindex = start_page->index + 1; tindex < tloff; tindex++) {
			page = xfs_probe_unwritten_page(mapping,
						tindex, iomapp, ioend,
						PAGE_CACHE_SIZE, &bs, bbits);
			if (!page)
				break;
			nblocks += bs;
			atomic_add(bs, &ioend->io_remaining);
			xfs_convert_page(inode, page, iomapp, wbc, ioend,
							startio, all_bh);
			/* stop if converting the next page might add
			 * enough blocks that the corresponding byte
			 * count won't fit in our ulong page buf length */
			if (nblocks >= ((ULONG_MAX - PAGE_SIZE) >> block_bits))
				goto enough;
		}

		if (tindex == tlast &&
		    (pg_offset = (i_size_read(inode) & (PAGE_CACHE_SIZE - 1)))) {
			page = xfs_probe_unwritten_page(mapping,
							tindex, iomapp, ioend,
							pg_offset, &bs, bbits);
			if (page) {
				nblocks += bs;
				atomic_add(bs, &ioend->io_remaining);
				xfs_convert_page(inode, page, iomapp, wbc, ioend,
							startio, all_bh);
				if (nblocks >= ((ULONG_MAX - PAGE_SIZE) >> block_bits))
					goto enough;
			}
		}
	}

enough:
	ioend->io_size = (xfs_off_t)nblocks << block_bits;
	ioend->io_offset = offset;
	xfs_finish_ioend(ioend);
	return 0;
}

STATIC void
xfs_submit_page(
	struct page		*page,
	struct writeback_control *wbc,
	struct buffer_head	*bh_arr[],
	int			bh_count,
	int			probed_page,
	int			clear_dirty)
{
	struct buffer_head	*bh;
	int			i;

	BUG_ON(PageWriteback(page));
	if (bh_count)
		set_page_writeback(page);
	if (clear_dirty)
		clear_page_dirty(page);
	unlock_page(page);

	if (bh_count) {
		for (i = 0; i < bh_count; i++) {
			bh = bh_arr[i];
			mark_buffer_async_write(bh);
			if (buffer_unwritten(bh))
				set_buffer_unwritten_io(bh);
			set_buffer_uptodate(bh);
			clear_buffer_dirty(bh);
		}

		for (i = 0; i < bh_count; i++)
			submit_bh(WRITE, bh_arr[i]);

		if (probed_page && clear_dirty)
			wbc->nr_to_write--;	/* Wrote an "extra" page */
	}
}

/*
 * Allocate & map buffers for page given the extent map. Write it out.
 * except for the original page of a writepage, this is called on
 * delalloc/unwritten pages only, for the original page it is possible
 * that the page has no mapping at all.
 */
STATIC void
xfs_convert_page(
	struct inode		*inode,
	struct page		*page,
	xfs_iomap_t		*iomapp,
	struct writeback_control *wbc,
	void			*private,
	int			startio,
	int			all_bh)
{
	struct buffer_head	*bh_arr[MAX_BUF_PER_PAGE], *bh, *head;
	xfs_iomap_t		*mp = iomapp, *tmp;
	unsigned long		offset, end_offset;
	int			index = 0;
	int			bbits = inode->i_blkbits;
	int			len, page_dirty;

	end_offset = (i_size_read(inode) & (PAGE_CACHE_SIZE - 1));

	/*
	 * page_dirty is initially a count of buffers on the page before
	 * EOF and is decrememted as we move each into a cleanable state.
	 */
	len = 1 << inode->i_blkbits;
	end_offset = max(end_offset, PAGE_CACHE_SIZE);
	end_offset = roundup(end_offset, len);
	page_dirty = end_offset / len;

	offset = 0;
	bh = head = page_buffers(page);
	do {
		if (offset >= end_offset)
			break;
		if (!(PageUptodate(page) || buffer_uptodate(bh)))
			continue;
		if (buffer_mapped(bh) && all_bh &&
		    !(buffer_unwritten(bh) || buffer_delay(bh))) {
			if (startio) {
				lock_buffer(bh);
				bh_arr[index++] = bh;
				page_dirty--;
			}
			continue;
		}
		tmp = xfs_offset_to_map(page, mp, offset);
		if (!tmp)
			continue;
		ASSERT(!(tmp->iomap_flags & IOMAP_HOLE));
		ASSERT(!(tmp->iomap_flags & IOMAP_DELAY));

		/* If this is a new unwritten extent buffer (i.e. one
		 * that we haven't passed in private data for, we must
		 * now map this buffer too.
		 */
		if (buffer_unwritten(bh) && !bh->b_end_io) {
			ASSERT(tmp->iomap_flags & IOMAP_UNWRITTEN);
			xfs_map_unwritten(inode, page, head, bh, offset,
					bbits, tmp, wbc, startio, all_bh);
		} else if (! (buffer_unwritten(bh) && buffer_locked(bh))) {
			xfs_map_at_offset(page, bh, offset, bbits, tmp);
			if (buffer_unwritten(bh)) {
				set_buffer_unwritten_io(bh);
				bh->b_private = private;
				ASSERT(private);
			}
		}
		if (startio) {
			bh_arr[index++] = bh;
		} else {
			set_buffer_dirty(bh);
			unlock_buffer(bh);
			mark_buffer_dirty(bh);
		}
		page_dirty--;
	} while (offset += len, (bh = bh->b_this_page) != head);

	if (startio && index) {
		xfs_submit_page(page, wbc, bh_arr, index, 1, !page_dirty);
	} else {
		unlock_page(page);
	}
}

/*
 * Convert & write out a cluster of pages in the same extent as defined
 * by mp and following the start page.
 */
STATIC void
xfs_cluster_write(
	struct inode		*inode,
	pgoff_t			tindex,
	xfs_iomap_t		*iomapp,
	struct writeback_control *wbc,
	int			startio,
	int			all_bh,
	pgoff_t			tlast)
{
	struct page		*page;

	for (; tindex <= tlast; tindex++) {
		page = xfs_probe_delalloc_page(inode, tindex);
		if (!page)
			break;
		xfs_convert_page(inode, page, iomapp, wbc, NULL,
				startio, all_bh);
	}
}

/*
 * Calling this without startio set means we are being asked to make a dirty
 * page ready for freeing it's buffers.  When called with startio set then
 * we are coming from writepage.
 *
 * When called with startio set it is important that we write the WHOLE
 * page if possible.
 * The bh->b_state's cannot know if any of the blocks or which block for
 * that matter are dirty due to mmap writes, and therefore bh uptodate is
 * only vaild if the page itself isn't completely uptodate.  Some layers
 * may clear the page dirty flag prior to calling write page, under the
 * assumption the entire page will be written out; by not writing out the
 * whole page the page can be reused before all valid dirty data is
 * written out.  Note: in the case of a page that has been dirty'd by
 * mapwrite and but partially setup by block_prepare_write the
 * bh->b_states's will not agree and only ones setup by BPW/BCW will have
 * valid state, thus the whole page must be written out thing.
 */

STATIC int
xfs_page_state_convert(
	struct inode	*inode,
	struct page	*page,
	struct writeback_control *wbc,
	int		startio,
	int		unmapped) /* also implies page uptodate */
{
	struct buffer_head	*bh_arr[MAX_BUF_PER_PAGE], *bh, *head;
	xfs_iomap_t		*iomp, iomap;
	loff_t			offset;
	unsigned long           p_offset = 0;
	__uint64_t              end_offset;
	pgoff_t                 end_index, last_index, tlast;
	int			len, err, i, cnt = 0, uptodate = 1;
	int			flags;
	int			page_dirty;

	/* wait for other IO threads? */
	flags = (startio && wbc->sync_mode != WB_SYNC_NONE) ? 0 : BMAPI_TRYLOCK;

	/* Is this page beyond the end of the file? */
	offset = i_size_read(inode);
	end_index = offset >> PAGE_CACHE_SHIFT;
	last_index = (offset - 1) >> PAGE_CACHE_SHIFT;
	if (page->index >= end_index) {
		if ((page->index >= end_index + 1) ||
		    !(i_size_read(inode) & (PAGE_CACHE_SIZE - 1))) {
			if (startio)
				unlock_page(page);
			return 0;
		}
	}

	end_offset = min_t(unsigned long long,
			(loff_t)(page->index + 1) << PAGE_CACHE_SHIFT, offset);
	offset = (loff_t)page->index << PAGE_CACHE_SHIFT;

	/*
	 * page_dirty is initially a count of buffers on the page before
	 * EOF and is decrememted as we move each into a cleanable state.
	 */
	len = 1 << inode->i_blkbits;
	p_offset = max(p_offset, PAGE_CACHE_SIZE);
	p_offset = roundup(p_offset, len);
	page_dirty = p_offset / len;

	iomp = NULL;
	p_offset = 0;
	bh = head = page_buffers(page);

	do {
		if (offset >= end_offset)
			break;
		if (!buffer_uptodate(bh))
			uptodate = 0;
		if (!(PageUptodate(page) || buffer_uptodate(bh)) && !startio)
			continue;

		if (iomp) {
			iomp = xfs_offset_to_map(page, &iomap, p_offset);
		}

		/*
		 * First case, map an unwritten extent and prepare for
		 * extent state conversion transaction on completion.
		 */
		if (buffer_unwritten(bh)) {
			if (!startio)
				continue;
			if (!iomp) {
				err = xfs_map_blocks(inode, offset, len, &iomap,
						BMAPI_WRITE|BMAPI_IGNSTATE);
				if (err) {
					goto error;
				}
				iomp = xfs_offset_to_map(page, &iomap,
								p_offset);
			}
			if (iomp) {
				if (!bh->b_end_io) {
					err = xfs_map_unwritten(inode, page,
							head, bh, p_offset,
							inode->i_blkbits, iomp,
							wbc, startio, unmapped);
					if (err) {
						goto error;
					}
				} else {
					set_bit(BH_Lock, &bh->b_state);
				}
				BUG_ON(!buffer_locked(bh));
				bh_arr[cnt++] = bh;
				page_dirty--;
			}
		/*
		 * Second case, allocate space for a delalloc buffer.
		 * We can return EAGAIN here in the release page case.
		 */
		} else if (buffer_delay(bh)) {
			if (!iomp) {
				err = xfs_map_blocks(inode, offset, len, &iomap,
						BMAPI_ALLOCATE | flags);
				if (err) {
					goto error;
				}
				iomp = xfs_offset_to_map(page, &iomap,
								p_offset);
			}
			if (iomp) {
				xfs_map_at_offset(page, bh, p_offset,
						inode->i_blkbits, iomp);
				if (startio) {
					bh_arr[cnt++] = bh;
				} else {
					set_buffer_dirty(bh);
					unlock_buffer(bh);
					mark_buffer_dirty(bh);
				}
				page_dirty--;
			}
		} else if ((buffer_uptodate(bh) || PageUptodate(page)) &&
			   (unmapped || startio)) {

			if (!buffer_mapped(bh)) {
				int	size;

				/*
				 * Getting here implies an unmapped buffer
				 * was found, and we are in a path where we
				 * need to write the whole page out.
				 */
				if (!iomp) {
					size = xfs_probe_unmapped_cluster(
							inode, page, bh, head);
					err = xfs_map_blocks(inode, offset,
							size, &iomap,
							BMAPI_WRITE|BMAPI_MMAP);
					if (err) {
						goto error;
					}
					iomp = xfs_offset_to_map(page, &iomap,
								     p_offset);
				}
				if (iomp) {
					xfs_map_at_offset(page,
							bh, p_offset,
							inode->i_blkbits, iomp);
					if (startio) {
						bh_arr[cnt++] = bh;
					} else {
						set_buffer_dirty(bh);
						unlock_buffer(bh);
						mark_buffer_dirty(bh);
					}
					page_dirty--;
				}
			} else if (startio) {
				if (buffer_uptodate(bh) &&
				    !test_and_set_bit(BH_Lock, &bh->b_state)) {
					bh_arr[cnt++] = bh;
					page_dirty--;
				}
			}
		}
	} while (offset += len, p_offset += len,
		((bh = bh->b_this_page) != head));

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

	if (startio) {
		xfs_submit_page(page, wbc, bh_arr, cnt, 0, !page_dirty);
	}

	if (iomp) {
		offset = (iomp->iomap_offset + iomp->iomap_bsize - 1) >>
					PAGE_CACHE_SHIFT;
		tlast = min_t(pgoff_t, offset, last_index);
		xfs_cluster_write(inode, page->index + 1, iomp, wbc,