From owner-freebsd-virtualization@freebsd.org Tue Mar 21 19:04:41 2017 Return-Path: Delivered-To: freebsd-virtualization@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id 3BC49D16546 for ; Tue, 21 Mar 2017 19:04:41 +0000 (UTC) (envelope-from bugzilla-noreply@freebsd.org) Received: from kenobi.freebsd.org (kenobi.freebsd.org [IPv6:2001:1900:2254:206a::16:76]) (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 10BDE830 for ; Tue, 21 Mar 2017 19:04:41 +0000 (UTC) (envelope-from bugzilla-noreply@freebsd.org) Received: from bugs.freebsd.org ([127.0.1.118]) by kenobi.freebsd.org (8.15.2/8.15.2) with ESMTP id v2LJ4eOd084066 for ; Tue, 21 Mar 2017 19:04:40 GMT (envelope-from bugzilla-noreply@freebsd.org) From: bugzilla-noreply@freebsd.org To: freebsd-virtualization@FreeBSD.org Subject: [Bug 216759] [kern] Memory speed with small blocks (1K) up to 35 times slower than host system (under QEMU emulation, but not only) Date: Tue, 21 Mar 2017 19:04:40 +0000 X-Bugzilla-Reason: AssignedTo X-Bugzilla-Type: changed X-Bugzilla-Watch-Reason: None X-Bugzilla-Product: Base System X-Bugzilla-Component: kern X-Bugzilla-Version: 11.0-RELEASE X-Bugzilla-Keywords: X-Bugzilla-Severity: Affects Many People X-Bugzilla-Who: andrew@azar-a.net X-Bugzilla-Status: New X-Bugzilla-Resolution: X-Bugzilla-Priority: --- X-Bugzilla-Assigned-To: freebsd-virtualization@FreeBSD.org X-Bugzilla-Flags: X-Bugzilla-Changed-Fields: Message-ID: In-Reply-To: References: Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Bugzilla-URL: https://bugs.freebsd.org/bugzilla/ Auto-Submitted: auto-generated MIME-Version: 1.0 X-BeenThere: freebsd-virtualization@freebsd.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: "Discussion of various virtualization techniques FreeBSD supports." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 21 Mar 2017 19:04:41 -0000 https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D216759 --- Comment #17 from andrew@azar-a.net --- (In reply to deJong from comment #16) It doesn't seem like the patchset is applied at all. > kern.timecounter.choice: TSC-low(-100) i8254(0) ACPI-fast(900) HPET(950) = dummy(-1000000) It should add kvmclock to that list which is obviously not there. Forcing TSC-low on KVM VM when the host trusts it more than HPET and ACPI-P= M is foolproof imo. It's obvious that the choice for this has been designed a long time ago and= has been programmed this way because back then TSC really sucked and any hyperv= isor except for programmed by VMWare loophole is penalised (don't know though wh= ich because xen requires more workarounds in the kernel and I don't really know= how bhyve runs it).. So obviously this part of code should be reworked if ((!smp_tsc && !tsc_is_invariant) || vm_guest) return (-100); X86_FEATURE_CONSTANT_TSC is not checked at all (this is from linux, but sti= ll BSD does not check that in any case in its TSC code (or I might be misreadi= ng the magic tsc_is_invariant variable)). And most of all the vm_guest check should go away. It's a terrible hack nowadays as this bug shows. I did not try this yet but maybe something like this will help if ((!smp_tsc && !tsc_is_invariant) || (vm_guest && vm_guest !=3D VM_GUEST_= KVM)) return (-100); --=20 You are receiving this mail because: You are the assignee for the bug.=