From owner-freebsd-current@freebsd.org Fri Apr 2 20:50:19 2021 Return-Path: Delivered-To: freebsd-current@mailman.nyi.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.nyi.freebsd.org (Postfix) with ESMTP id 6D97D5B1B29; Fri, 2 Apr 2021 20:50:19 +0000 (UTC) (envelope-from mchouque@thi.eu.com) Received: from relay5-d.mail.gandi.net (relay5-d.mail.gandi.net [217.70.183.197]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 4FBsd70G3Lz3tN5; Fri, 2 Apr 2021 20:50:18 +0000 (UTC) (envelope-from mchouque@thi.eu.com) X-Originating-IP: 62.210.143.248 Received: from weirdfishes.localdomain (62-210-143-248.rev.poneytelecom.eu [62.210.143.248]) (Authenticated sender: m@thi.eu.com) by relay5-d.mail.gandi.net (Postfix) with ESMTPSA id 663491C0003; Fri, 2 Apr 2021 20:50:15 +0000 (UTC) Received: by weirdfishes.localdomain (Postfix, from userid 1000) id 278907203B9D5; Fri, 2 Apr 2021 22:50:15 +0200 (CEST) Date: Fri, 2 Apr 2021 22:50:15 +0200 From: Mathieu Chouquet-Stringer To: Dave Cottlehuber Cc: freebsd-fs , freebsd-current , bryanv@freebsd.org Subject: Re: Scrub incredibly slow with 13.0-RC3 (as well as RC1 & 2) Message-ID: References: <3c00c109-ed5a-496c-9339-22b3dc31747a@www.fastmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3c00c109-ed5a-496c-9339-22b3dc31747a@www.fastmail.com> X-Face: %JOeya=Dg!}[/#Go&*&cQ+)){p1c8}u\Fg2Q3&)kothIq|JnWoVzJtCFo~4X List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 02 Apr 2021 20:50:19 -0000 On Fri, Apr 02, 2021 at 08:29:43PM +0000, Dave Cottlehuber wrote: > Does the issue also go away if you use `TSC-slow` ? I haven't, I can try it out. Yeah TSC-slow is poorly rated (-100) but scrubbing my zroot is much faster with it compared to ACPI-fast (rated at 900).. TSC-slow: scan: scrub repaired 0B in 00:00:12 with 0 errors on Fri Apr 2 22:37:06 2021 scan: scrub repaired 0B in 00:00:09 with 0 errors on Fri Apr 2 22:41:17 2021 scan: scrub repaired 0B in 00:00:09 with 0 errors on Fri Apr 2 22:41:40 2021 ACPI-fast: scan: scrub repaired 0B in 00:03:33 with 0 errors on Fri Apr 2 22:40:47 2021 scan: scrub repaired 0B in 00:03:29 with 0 errors on Fri Apr 2 22:46:14 2021 scan: scrub repaired 0B in 00:03:38 with 0 errors on Fri Apr 2 22:49:52 2021 So really ACPI-fast is not appropriately named, at least under KVM. Same thing as HPET... TSC-slow looks as good as kvmclock based on that simple benchmark but I guess there's a reason why it's at the bottom of the pile. > I investigated exactly this issue yesterday[1], and it's great to see Bryan's patch > resurrected, I will try it here and report back too. > > [1]: https://hackmd.io/HhYlbsDJTpCWHSwlrwrY_w FYI I created a new revision to have that patch merged (and there's a better version of it there): https://reviews.freebsd.org/D29531 Honestly the kvmclock patch is a game changer but I look forward to hearing your feedback. -- Mathieu Chouquet-Stringer The sun itself sees not till heaven clears. -- William Shakespeare --