diff options
| author | Jaegeuk Kim <jaegeuk.kim@samsung.com> | 2014-04-30 09:22:45 +0900 |
|---|---|---|
| committer | Jaegeuk Kim <jaegeuk.kim@samsung.com> | 2014-05-07 10:21:58 +0900 |
| commit | d5f66990bb928e7490ba4da94d585f618adcee5e (patch) | |
| tree | 41dd37a53c947a47d6ec4c0e2c648bd1268fae0b /fs/f2fs | |
| parent | bde446866c9200fcfa27ccde213d83db9a6827db (diff) | |
| download | linux-d5f66990bb928e7490ba4da94d585f618adcee5e.tar.gz | |
f2fs: decrease the lock granularity during write_begin
This patch reduces the lock granularity during write_begin. When the system is under memory pressure, it would be better to reduce the locking time for the data pages. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
Diffstat (limited to 'fs/f2fs')
| -rw-r--r-- | fs/f2fs/data.c | 14 |
1 files changed, 13 insertions, 1 deletions
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 91ff104b3849..a7bb98f5831a 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -909,6 +909,10 @@ repeat: page = grab_cache_page_write_begin(mapping, index, flags); if (!page) return -ENOMEM; + + /* to avoid latency during memory pressure */ + unlock_page(page); + *pagep = page; if (f2fs_has_inline_data(inode) && (pos + len) <= MAX_INLINE_DATA) @@ -920,10 +924,18 @@ repeat: f2fs_unlock_op(sbi); if (err) { - f2fs_put_page(page, 1); + f2fs_put_page(page, 0); return err; } inline_data: + lock_page(page); + if (unlikely(page->mapping != mapping)) { + f2fs_put_page(page, 1); + goto repeat; + } + + f2fs_wait_on_page_writeback(page, DATA); + if ((len == PAGE_CACHE_SIZE) || PageUptodate(page)) return 0; |