From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:45849) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1bwVYh-0008Dm-Dr for qemu-devel@nongnu.org; Tue, 18 Oct 2016 10:40:04 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1bwVYe-00087o-7q for qemu-devel@nongnu.org; Tue, 18 Oct 2016 10:40:03 -0400 Received: from mx1.redhat.com ([209.132.183.28]:53082) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1bwVYd-000876-WD for qemu-devel@nongnu.org; Tue, 18 Oct 2016 10:40:00 -0400 Date: Tue, 18 Oct 2016 16:39:55 +0200 From: Igor Mammedov Message-ID: <20161018163955.524972fd@nial.brq.redhat.com> In-Reply-To: <20161018125557.GN3275@thinpad.lan.raisama.net> References: <20161013135634.GB15008@potion> <1476444335-206575-1-git-send-email-imammedo@redhat.com> <20161018112737.GK3275@thinpad.lan.raisama.net> <20161018144455.3da0e986@nial.brq.redhat.com> <20161018125557.GN3275@thinpad.lan.raisama.net> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH v4 13/13] pc: require IRQ remapping and EIM if there could be x2APIC CPUs List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Eduardo Habkost Cc: qemu-devel@nongnu.org, rkrcmar@redhat.com, liuxiaojian6@huawei.com, mst@redhat.com, peterx@redhat.com, kevin@koconnor.net, kraxel@redhat.com, pbonzini@redhat.com, lersek@redhat.com, chao.gao@intel.com On Tue, 18 Oct 2016 10:55:57 -0200 Eduardo Habkost wrote: > On Tue, Oct 18, 2016 at 02:44:55PM +0200, Igor Mammedov wrote: > > On Tue, 18 Oct 2016 09:27:37 -0200 > > Eduardo Habkost wrote: > > =20 > > > On Fri, Oct 14, 2016 at 01:25:35PM +0200, Igor Mammedov wrote: =20 > > > > it would prevent starting guest with incorrect configs > > > > where interrupts couldn't be delivered to CPUs with > > > > APIC IDs > 255. > > > >=20 > > > > Signed-off-by: Igor Mammedov > > > > Reviewed-by: Radim Kr=C4=8Dm=C3=A1=C5=99 > > > > --- > > > > v4: > > > > - s/254/255/ in commit message (Radim) > > > > --- > > > > hw/i386/pc.c | 13 +++++++++++++ > > > > 1 file changed, 13 insertions(+) > > > >=20 > > > > diff --git a/hw/i386/pc.c b/hw/i386/pc.c > > > > index 40eb43b..f7070e0 100644 > > > > --- a/hw/i386/pc.c > > > > +++ b/hw/i386/pc.c > > > > @@ -68,6 +68,7 @@ > > > > #include "qapi-visit.h" > > > > #include "qom/cpu.h" > > > > #include "hw/nmi.h" > > > > +#include "hw/i386/intel_iommu.h" > > > > =20 > > > > /* debug PC/ISA interrupts */ > > > > //#define DEBUG_IRQ > > > > @@ -1264,6 +1265,18 @@ void pc_machine_done(Notifier *notifier, voi= d *data) > > > > sizeof(pcms->boot_cpus_le)); > > > > } > > > > } > > > > + > > > > + if (pcms->apic_id_limit > 255) { > > > > + IntelIOMMUState *iommu =3D INTEL_IOMMU_DEVICE(x86_iommu_ge= t_default()); > > > > + > > > > + if (!iommu || !iommu->x86_iommu.intr_supported || > > > > + iommu->intr_eim !=3D ON_OFF_AUTO_ON) { > > > > + error_report("current -smp configuration requires " > > > > + "Extended Interrupt Mode enabled. " > > > > + "IOMMU should have eim=3Don option set");= =20 > > >=20 > > > Suggestion for a follow-up patch: > > >=20 > > > * Error message explaining how to set eim=3Don if the iommu is > > > available > > > * Error message explaining how to make sure the iommu is created, > > > in case it was not even created. =20 > > Reason I didn't include how to create iommu/CLI example is that > > it could be some other iommu in future so that message could > > bit rot over time. =20 >=20 > I see. >=20 > >=20 > > But I can add description if you'd prefer it. > > How about something like this: > > + error_report("current -smp configuration requires " > > + "Intel IOMMU with Extended Interrupt Mode ena= bled. " > > + "To enable IOMMU add to command line: " > > + "-device intel-iommu,intremap=3Don,eim=3Don")= ; =20 >=20 > If there could be other iommu devices in the future, we can show > it as an example, but not an instruction to be followed as-is. >=20 > Maybe we can just say "You can add an IOMMU using: > -device intel-iommu,intremap=3Don,eim=3Don". It would mean it is one > way to add an IOMMU, but not necessarily the only way. ok, I'll amend error message. >=20 > BTW, are there any plans to make machine code create an IOMMU > automatically if the VM config requires it? not that I know of