From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:46205) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1g9u3D-0003VE-Ss for qemu-devel@nongnu.org; Tue, 09 Oct 2018 11:36:00 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1g9u38-0002qy-Pc for qemu-devel@nongnu.org; Tue, 09 Oct 2018 11:35:59 -0400 Received: from mx1.redhat.com ([209.132.183.28]:49002) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1g9u38-0002qa-Fw for qemu-devel@nongnu.org; Tue, 09 Oct 2018 11:35:54 -0400 References: <1538748792-19444-1-git-send-email-thuth@redhat.com> <875zybzdzm.fsf@dusky.pond.sub.org> <87zhvntg6k.fsf@dusky.pond.sub.org> <20181009154312.771c276a.cohuck@redhat.com> <20181009142309.GH22838@redhat.com> From: Paolo Bonzini Message-ID: <7d69f6c1-e493-8013-3045-fe71ffa0f283@redhat.com> Date: Tue, 9 Oct 2018 17:35:44 +0200 MIME-Version: 1.0 In-Reply-To: <20181009142309.GH22838@redhat.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH] accel: Improve selection of the default accelerator List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: =?UTF-8?Q?Daniel_P=2e_Berrang=c3=a9?= , Peter Maydell Cc: Cornelia Huck , Thomas Huth , Markus Armbruster , QEMU Developers On 09/10/2018 16:23, Daniel P. Berrang=C3=A9 wrote: >>> >>> (And I'd prefer kvm to be first in that list; anything that relies on >>> tcg being used should specify it explicitly... a normal user will >>> likely always want the fast variant.) >> tcg should be the default for binaries without KVM compiled in, >> of course... But as Thomas points out, the reason for our current >> default is the usual "because we tend not to change things that >> would break existing working command lines". ... and because we rely on people that want KVM using almost always the distro qemu-kvm. > Putting KVM first shouldn't break existing working command lines > in general. >=20 > If user doesn't have KVM it won't have any impact as it'll still > fallback to TCG. If user does have KVM, their VMs will be faster.=20 >=20 > It could conceivably break if >=20 > - TCG had some different behaviour to KVM that > the guest relied upon (we should fix such differences) TCG can emulate stuff that doesn't exist in the host processor (e.g. MPX, PKRU, and on older processors SMEP/SMAP though those should be ubiquitous). There are (or were) CI systems for the kernel that were using it that way. Paolo > - KVM kmod was buggy (eg broken nested virt, again should be > something we fix) > - If the guest OS relied on things being slow (that's already > fragile) >=20 > IMHO we should assume that TCG & KVM are functionally equivalent > and non-broken, and thus changing default accelerator should not > be considered an incompatible change.