Btrfs: don't call writepages from within write_full_page (0d10ee2e) · Commits · 戴 / test

When doing a writepage we call writepages to try and write out any other dirty pages in the area. This could cause problems where we commit a transaction and then have somebody else dirtying metadata in the area as we could end up writing out a lot more than we care about, which could cause latency on anybody who is waiting for the transaction to completely finish committing. Thanks, Signed-off-by:

Josef Bacik <josef@redhat.com> Signed-off-by:

Chris Mason <chris.mason@oracle.com>

fs/btrfs/extent_io.c

+0 −10

Original line number	Diff line number	Diff line
		@@ -2502,7 +2502,6 @@ int extent_write_full_page(struct extent_io_tree tree, struct page page,
		struct writeback_control *wbc)
		{
		int ret;
		struct address_space *mapping = page->mapping;
		struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		@@ -2510,18 +2509,9 @@ int extent_write_full_page(struct extent_io_tree tree, struct page page,
		.extent_locked = 0,
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
		};
		struct writeback_control wbc_writepages = {
		.sync_mode = wbc->sync_mode,
		.older_than_this = NULL,
		.nr_to_write = 64,
		.range_start = page_offset(page) + PAGE_CACHE_SIZE,
		.range_end = (loff_t)-1,
		};

		ret = __extent_writepage(page, wbc, &epd);

		extent_write_cache_pages(tree, mapping, &wbc_writepages,
		__extent_writepage, &epd, flush_write_bio);
		flush_epd_write_bio(&epd);
		return ret;
		}

Admin message