On Mon, Jun 15, 2020 at 10:38:44AM +0100, fdmanana@xxxxxxxxxx wrote:
> From: Filipe Manana <fdmanana@xxxxxxxx>
>
> This brings back an optimization that commit e678934cbe5f02 ("btrfs:
> Remove unnecessary check from join_running_log_trans") removed, but in
> a different form. So it's almost equivalent to a revert.
I very much prefer the bit to be the synchronization mechanism, the
logic is easy to follow instead of the cryptic barrier.
The original patch came with numbers to support the 'not needed and no
perf impact
(https://lore.kernel.org/linux-btrfs/20190523115126.10532-1-nborisov@xxxxxxxx/)
but it probably wasn't triggering the right load.
[...]
> The test robots from intel reported a -30.7% performance regression for
> a REAIM test after commit e678934cbe5f02 ("btrfs: Remove unnecessary check
> from join_running_log_trans").
Thanks for fixing the perf regression and points for the test robot too.
