diff options
| author |   Chao Yu <yuchao0@huawei.com> | 2018-06-04 23:20:36 +0800 |
|---|---|---|
| committer |   Jaegeuk Kim <jaegeuk@kernel.org> | 2018-06-04 14:33:20 -0700 |
| commit | c29fd0c0e26dacb7a33ad166587059818a94b4e0 (patch) | |
| tree | 6a25f17833047796fccde3cd7c4e17d35199e640 /fs/f2fs/gc.c | |
| parent | f2fs: don't change wbc->sync_mode (diff) | |
| download | linux-dev-c29fd0c0e26dacb7a33ad166587059818a94b4e0.tar.xz linux-dev-c29fd0c0e26dacb7a33ad166587059818a94b4e0.zip | |
f2fs: let sync node IO interrupt async one
Although mixed sync/async IOs can have continuous LBA, as they have
different IO priority, block IO scheduler will add them into different
queues and commit them separately, result in splited IOs which causes
wrose performance.
This patch gives high priority to synchronous IO of nodes, means that
once synchronous flow starts, it can interrupt asynchronous writeback
flow of system flusher, so more big IOs can be expected.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Diffstat (limited to 'fs/f2fs/gc.c')
| -rw-r--r-- | fs/f2fs/gc.c | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c index 2a97ce7bc91d..9093be6e7a7d 100644 --- a/fs/f2fs/gc.c +++ b/fs/f2fs/gc.c @@ -473,12 +473,16 @@ static void gc_node_segment(struct f2fs_sb_info *sbi, block_t start_addr; int off; int phase = 0; + bool fggc = (gc_type == FG_GC); start_addr = START_BLOCK(sbi, segno); next_step: entry = sum; + if (fggc && phase == 2) + atomic_inc(&sbi->wb_sync_req[NODE]); + for (off = 0; off < sbi->blocks_per_seg; off++, entry++) { nid_t nid = le32_to_cpu(entry->nid); struct page *node_page; @@ -525,6 +529,9 @@ next_step: if (++phase < 3) goto next_step; + + if (fggc) + atomic_dec(&sbi->wb_sync_req[NODE]); } /* |