Re: btrfs write-bandwidth performance regression of 6.5-rc4/rc3

[Wang Yugui <wangyugui@e16-tech.com>]

Hi,

> On Tue, Aug 01, 2023 at 09:04:05PM +0800, Wang Yugui wrote:
> > good performance
> > 	drop 'btrfs: submit IO synchronously for fast checksum implementations' too
> > 	6.4.0 + patches until ' btrfs: use SECTOR_SHIFT to convert LBA to physical offset'
> > 
> > but I have tested 6.1.y with  a patch almost same as 
> > 'btrfs: submit IO synchronously for fast checksum implementations'
> > for over 20+ times, no performance regression found.
> 
> Can you try a git-revert of 140fb1f734736a on the latest tree (which
> should work cleanly) for an additional data point?

GOOD performance  when btrfs 6.5-rc4 with
	Revert "btrfs: determine synchronous writers from bio or writeback control"
	Revert "btrfs: submit IO synchronously for fast checksum implementations"

GOOD performance  when btrfs 6.5-rc4 with this fix too.
diff --git a/fs/btrfs/bio.c b/fs/btrfs/bio.c
index 1f3e06ec6924..1b7344e673db 100644
--- a/fs/btrfs/bio.c
+++ b/fs/btrfs/bio.c
@@ -598,7 +598,7 @@ static void run_one_async_free(struct btrfs_work *work)
 static bool should_async_write(struct btrfs_bio *bbio)
 {
        /* Submit synchronously if the checksum implementation is fast. */
-       if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &bbio->fs_info->flags))
+       if ((bbio->bio.bi_opf & REQ_META) && test_bit(BTRFS_FS_CSUM_IMPL_FAST, &bbio->fs_info->flags))
                return false;

        /*

Best Regards
Wang Yugui (wangyugui@e16-tech.com)
2023/08/01