[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <aBJv7AuXubBDQNIq@slm.duckdns.org>
Date: Wed, 30 Apr 2025 08:46:04 -1000
From: Tejun Heo <tj@...nel.org>
To: Junxuan Liao <ljx@...wisc.edu>
Cc: Chris Mason <clm@...com>, Josef Bacik <josef@...icpanda.com>,
David Sterba <dsterba@...e.com>,
Lai Jiangshan <jiangshanlai@...il.com>, linux-btrfs@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: btrfs thread_pool_size logic out of sync with workqueue
On Tue, Apr 29, 2025 at 04:26:52PM -0500, Junxuan Liao wrote:
> On 4/29/25 4:15 PM, Junxuan Liao wrote:
> > So versions from 6.6 to 6.8 do have the bug, right? I guess the
> > performance regression isn't easy to trigger so no one noticed and
> > reported it.
>
> My bad. I missed that it has already been reported in 2023.
> https://lore.kernel.org/all/dbu6wiwu3sdhmhikb2w6lns7b27gbobfavhjj57kwi2quafgwl@htjcc5oikcr3/
Yeah, the changes were too invasive to backport through -stable especially
at the time as I didn't know how well the new scheme would work. There have
been some updates afterwards but it seemed to have held up fine, so if some
distros wanna backport them, this should be pretty safe now. However, I
think the window has already passed for -stable backports.
Thanks.
--
tejun
Powered by blists - more mailing lists