From 092c8d46e348b5fa4109a06d8a1246060e09dc8c Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Mon, 29 Apr 2013 15:06:17 -0700 Subject: direct-io: fix boundary block handling When we read/write a file sequentially, we will read/write not only the data blocks but also the indirect blocks that may not be physically adjacent to the data blocks. So filesystems set the BH_Boundary flag to submit the previous I/O before reading/writing an indirect block. However the generic direct IO code mishandles buffer_boundary(), setting sdio->boundary before each submit_page_section() call which results in sending only one page bios as underlying code thinks this page is the last in the contiguous extent. So fix the problem by setting sdio->boundary only if the current page is really the last one in the mapped extent. With this patch and "direct-io: submit bio after boundary buffer is added to it" I've measured about 10% throughput improvement of direct IO reads on ext3 with SATA harddrive (from 90 MB/s to 100 MB/s). With ramdisk, the improvement was about 3-fold (from 350 MB/s to 1.2 GB/s). For other filesystems (such as ext4), the improvements won't be as visible because the frequency of BH_Boundary flag being set is much smaller. Signed-off-by: Jan Kara Reported-by: Kazuya Mio Tested-by: Kazuya Mio Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/direct-io.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/direct-io.c b/fs/direct-io.c index f853263cf74f..9644d0205dad 100644 --- a/fs/direct-io.c +++ b/fs/direct-io.c @@ -969,7 +969,8 @@ do_holes: this_chunk_bytes = this_chunk_blocks << blkbits; BUG_ON(this_chunk_bytes == 0); - sdio->boundary = buffer_boundary(map_bh); + if (this_chunk_blocks == sdio->blocks_available) + sdio->boundary = buffer_boundary(map_bh); ret = submit_page_section(dio, sdio, page, offset_in_page, this_chunk_bytes, -- cgit v1.2.3