On 12/27/25 9:22 AM, Chao Yu via Linux-f2fs-devel wrote:
On 12/26/2025 11:39 AM, Yongpeng Yang wrote:
On 12/26/25 09:52, Chao Yu via Linux-f2fs-devel wrote:
On 12/25/2025 8:59 PM, Yongpeng Yang wrote:
From: Yongpeng Yang <[email protected]>
f2fs_folio_wait_writeback ensures the folio write is submitted to the
Actually, __submit_merged_write_cond() was called from many other
places,
for those scenario, we can not ensure whether the DATA or NODE type
pages
were cached in one bio cache or not, so we decide to submit all
cached bios
w/ DATA or NODE type.
block layer via __submit_merged_write_cond, then waits for the folio to
complete. Other I/O submissions are irrelevant to
f2fs_folio_wait_writeback. Thus, if the folio write bio is already
For f2fs_folio_wait_writeback() case, I guess you can optimize it
separately,
maybe passing a parameter to __submit_merged_write_cond() as a hint?
Currently, only the call chain f2fs_folio_wait_writeback
-> f2fs_submit_merged_write_cond->__submit_merged_write_cond passes a
non-NULL folio with a NULL inode. When the return value of
__has_merged_page is true and the folio is non-NULL, it indicates the
call is from f2fs_folio_wait_writeback. Indeed, this is unclear.
The force parameter seems redundant, where force = true implies `inode
== NULL && folio == NULL && ino == 0` is true, and force = false implies
`inode != NULL || folio != NULL || ino != 0` is true.
How about removing the force parameter and adding a `bool writeback`
parameter to indicate calls from f2fs_folio_wait_writeback?
Sure, can we clean up force parameter in separated patch?
OK, I'll clean it up and optmize f2fs_folio_wait_writeback in v2 patch.
Thanks
Yongpeng,
Thanks,
Thanks
Yongpeng,
Thanks,
submitted, the function can return immediately.
Signed-off-by: Yongpeng Yang <[email protected]>
---
fs/f2fs/data.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 12bf4b6e0075..c743352b686d 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -678,8 +678,12 @@ static void __submit_merged_write_cond(struct
f2fs_sb_info *sbi,
ret = __has_merged_page(io->bio, inode, folio, ino);
f2fs_up_read(&io->io_rwsem);
}
- if (ret)
+ if (ret) {
__f2fs_submit_merged_write(sbi, type, temp);
+ /* don't need to submit other types of bio. */
+ if (folio)
+ break;
+ }
/* TODO: use HOT temp only for meta pages now. */
if (type >= META)
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
