fs/mpage.c: forgotten WRITE_SYNC in case of WB_SYNC_ALL
PATCH v2 1/1 fs/mpage.c: forgotten WRITE_SYNC in case of WB_SYNC_ALL
When data integrity operation happens (sync, fsync, fdatasync calls) writeback control is set to WB_SYNC_ALL. In that case all write requests are marked with WRITE_SYNC (WRITE | REQ_SYNC | REQ_NOIDLE) indicating that caller is waiting for completion and block layer or block device should prioritize the IO avoiding any possible delays.
But mpage writeback path ignores marking requests as WRITE_SYNC.
This patch fixes this.
--- fs/mpage.c | 23 +++++++++++++++-------- 1 file changed, 15 insertions(+), 8 deletions(-)
fs/mpage.c +15 -8
@@ -462,6 +462,7 @@ static int __mpage_writepage(struct page *page, struct writeback_control *wbc,struct buffer_head map_bh;loff_t i_size =i_size_read (inode);int ret =0 ; +int wr = (wbc->sync_mode ==WB_SYNC_ALL ?WRITE_SYNC :WRITE );if (page_has_buffers (page)) {struct buffer_head *head =page_buffers (page);
@@ -570,7 +571,7 @@ page_is_mapped:* This page will go to BIO. Do we need to send this BIO off first? */ if (bio && mpd->last_block_in_bio != blocks[0 ] -1 ) - bio =mpage_bio_submit (WRITE , bio); + bio =mpage_bio_submit (wr, bio); alloc_new:if (bio ==NULL ) {
@@ -587,7 +588,7 @@ alloc_new:*/ length = first_unmapped << blkbits;if (bio_add_page (bio, page, length,0 ) < length) { - bio =mpage_bio_submit (WRITE , bio); + bio =mpage_bio_submit (wr, bio); goto alloc_new; }
@@ -620,7 +621,7 @@ alloc_new:set_page_writeback (page);unlock_page (page);if (boundary || (first_unmapped != blocks_per_page)) { - bio =mpage_bio_submit (WRITE , bio); + bio =mpage_bio_submit (wr, bio);if (boundary_block) {write_boundary_block (boundary_bdev, boundary_block,1 << blkbits);
@@ -632,7 +633,7 @@ alloc_new: confused:if (bio) - bio =mpage_bio_submit (WRITE , bio); + bio =mpage_bio_submit (wr, bio);if (mpd->use_writepage) { ret = mapping->a_ops->writepage (page, wbc);
@@ -688,8 +689,11 @@ mpage_writepages(struct address_space *mapping, }; ret =write_cache_pages (mapping, wbc, __mpage_writepage, &mpd); -if (mpd.bio) -mpage_bio_submit (WRITE , mpd.bio); +if (mpd.bio) { +int wr = (wbc->sync_mode ==WB_SYNC_ALL ? +WRITE_SYNC :WRITE ); +mpage_bio_submit (wr, mpd.bio); + } }blk_finish_plug (&plug);return ret;
@@ -706,8 +710,11 @@ int mpage_writepage(struct page *page, get_block_t get_block, .use_writepage = 0, };int ret =__mpage_writepage (page, wbc, &mpd); -if (mpd.bio) -mpage_bio_submit (WRITE , mpd.bio); +if (mpd.bio) { +int wr = (wbc->sync_mode ==WB_SYNC_ALL ? +WRITE_SYNC :WRITE ); +mpage_bio_submit (wr, mpd.bio); + }return ret; }EXPORT_SYMBOL (mpage_writepage);
Signature · 1 line
1.9.0
Quoted 9 lines · On Sat, 15 Mar 2014 00:34:28 +0900 Roman Pen <[email protected]> wrote:
When data integrity operation happens (sync, fsync, fdatasync calls) writeback control is set to
WB_SYNC_ALL. In that case all write requests are marked withWRITE_SYNC(WRITE |REQ_SYNC|REQ_NOIDLE) indicating that caller is waiting for completion and block layer or block device should prioritize the IO avoiding any possible delays.But mpage writeback path ignores marking requests as
WRITE_SYNC.This patch fixes this.
Sorry, but it would be unwise to merge this without a good understanding of its runtime effects.