From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-18.5 required=3.0 tests=BAYES_00,DKIM_INVALID, DKIM_SIGNED,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER,INCLUDES_PATCH, MAILING_LIST_MULTI,MENTIONS_GIT_HOSTING,SPF_HELO_NONE,SPF_PASS,URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 16923C433E0 for ; Mon, 21 Dec 2020 13:28:23 +0000 (UTC) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 4DF2322B51 for ; Mon, 21 Dec 2020 13:28:22 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 4DF2322B51 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=redhat.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Received: from localhost ([::1]:56384 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1krLEb-0005Gv-59 for qemu-devel@archiver.kernel.org; Mon, 21 Dec 2020 08:28:21 -0500 Received: from eggs.gnu.org ([2001:470:142:3::10]:53556) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1krLAz-0004Ca-Dk for qemu-devel@nongnu.org; Mon, 21 Dec 2020 08:24:37 -0500 Received: from us-smtp-delivery-124.mimecast.com ([63.128.21.124]:55409) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_CBC_SHA1:256) (Exim 4.90_1) (envelope-from ) id 1krLAu-0004fT-A0 for qemu-devel@nongnu.org; Mon, 21 Dec 2020 08:24:36 -0500 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1608557068; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=Ca03RvTNGVHMob98GXsi+/ZMYRTzgFfE+NY9/FeLQSQ=; b=QABICLm5cNQKAkVm+bPIu9ruU5Ck3afipgmmU5gg6TSAVQPIWvOelADe127gYQWSB4GksY krsK4qYxx9kHsn8YtJDC3vNgjGkPe4Ed6xHHt60PZiUWqEhKkS/udXHuIjT06V3qQkRoO6 fg1TzchMjbjbLPFCbIqLYtyvZfsVQ5o= Received: from mimecast-mx01.redhat.com (mimecast-mx01.redhat.com [209.132.183.4]) (Using TLS) by relay.mimecast.com with ESMTP id us-mta-226-Cvx_tR_ZM9CLbEAr8HleWA-1; Mon, 21 Dec 2020 08:24:26 -0500 X-MC-Unique: Cvx_tR_ZM9CLbEAr8HleWA-1 Received: from smtp.corp.redhat.com (int-mx04.intmail.prod.int.phx2.redhat.com [10.5.11.14]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx01.redhat.com (Postfix) with ESMTPS id 655651009475 for ; Mon, 21 Dec 2020 13:24:25 +0000 (UTC) Received: from localhost (ovpn-113-127.ams2.redhat.com [10.36.113.127]) by smtp.corp.redhat.com (Postfix) with ESMTP id 06F285D9CA; Mon, 21 Dec 2020 13:24:20 +0000 (UTC) Date: Mon, 21 Dec 2020 14:24:18 +0100 From: Igor Mammedov To: Eduardo Habkost Subject: Re: [PATCH 5/5] i386: provide simple 'hyperv=on' option to x86 machine types Message-ID: <20201221142418.0863bc59@redhat.com> In-Reply-To: <20201218180721.GS3140057@habkost.net> References: <20201119103221.1665171-1-vkuznets@redhat.com> <20201119103221.1665171-6-vkuznets@redhat.com> <20201216205202.GJ3140057@habkost.net> <20201218181340.5e398280@redhat.com> <20201218180721.GS3140057@habkost.net> MIME-Version: 1.0 X-Scanned-By: MIMEDefang 2.79 on 10.5.11.14 Authentication-Results: relay.mimecast.com; auth=pass smtp.auth=CUSA124A263 smtp.mailfrom=imammedo@redhat.com X-Mimecast-Spam-Score: 0 X-Mimecast-Originator: redhat.com Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Received-SPF: pass client-ip=63.128.21.124; envelope-from=imammedo@redhat.com; helo=us-smtp-delivery-124.mimecast.com X-Spam_score_int: -20 X-Spam_score: -2.1 X-Spam_bar: -- X-Spam_report: (-2.1 / 5.0 requ) BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_NONE=-0.0001, RCVD_IN_MSPIKE_H4=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Paolo Bonzini , Vitaly Kuznetsov , Marcelo Tosatti , qemu-devel@nongnu.org Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: "Qemu-devel" On Fri, 18 Dec 2020 13:07:21 -0500 Eduardo Habkost wrote: > On Fri, Dec 18, 2020 at 06:13:40PM +0100, Igor Mammedov wrote: > > On Wed, 16 Dec 2020 15:52:02 -0500 > > Eduardo Habkost wrote: > > > > > On Thu, Nov 19, 2020 at 11:32:21AM +0100, Vitaly Kuznetsov wrote: > > > > Enabling Hyper-V emulation for a Windows VM is a tiring experience as it > > > > requires listing all currently supported enlightenments ("hv_*" CPU > > > > features) explicitly. We do have a 'hv_passthrough' mode enabling > > > > everything but it can't be used in production as it prevents migration. > > > > > > > > Introduce a simple 'hyperv=on' option for all x86 machine types enabling > > > > all currently supported Hyper-V enlightenments. Later, when new > > > > enlightenments get implemented, we will be adding them to newer machine > > > > types only (by disabling them for legacy machine types) thus preserving > > > > migration. > > > > > > > > Signed-off-by: Vitaly Kuznetsov > > > > Reviewed-by: Eduardo Habkost > > > [...] > > > > --- > > > > docs/hyperv.txt | 8 ++++++++ > > > > hw/i386/x86.c | 30 ++++++++++++++++++++++++++++++ > > > > include/hw/i386/x86.h | 7 +++++++ > > > > target/i386/cpu.c | 14 ++++++++++++++ > > > > 4 files changed, 59 insertions(+) > > > > > > > > diff --git a/docs/hyperv.txt b/docs/hyperv.txt > > > > index 5df00da54fc4..1a76a07f8417 100644 > > > > --- a/docs/hyperv.txt > > > > +++ b/docs/hyperv.txt > > > > @@ -29,6 +29,14 @@ When any set of the Hyper-V enlightenments is enabled, QEMU changes hypervisor > > > > identification (CPUID 0x40000000..0x4000000A) to Hyper-V. KVM identification > > > > and features are kept in leaves 0x40000100..0x40000101. > > > > > > > > +Hyper-V enlightenments can be enabled in bulk by specifying 'hyperv=on' to an > > > > +x86 machine type: > > > > + > > > > + qemu-system-x86_64 -machine q35,accel=kvm,kernel-irqchip=split,hyperv=on ... > > > > + > > > > +Note, new enlightenments are only added to the latest (in-develompent) machine > > > > +type, older machine types keep the list of the supported features intact to > > > > +safeguard migration. > > > > > > > > 3. Existing enlightenments > > > > =========================== > > > > diff --git a/hw/i386/x86.c b/hw/i386/x86.c > > > > index 5944fc44edca..57f27d56ecc6 100644 > > > > --- a/hw/i386/x86.c > > > > +++ b/hw/i386/x86.c > > > > @@ -1171,6 +1171,20 @@ static void x86_machine_set_acpi(Object *obj, Visitor *v, const char *name, > > > > visit_type_OnOffAuto(v, name, &x86ms->acpi, errp); > > > > } > > > > > > > > +static bool x86_machine_get_hyperv(Object *obj, Error **errp) > > > > +{ > > > > + X86MachineState *x86ms = X86_MACHINE(obj); > > > > + > > > > + return x86ms->hyperv_enabled; > > > > +} > > > > + > > > > +static void x86_machine_set_hyperv(Object *obj, bool value, Error **errp) > > > > +{ > > > > + X86MachineState *x86ms = X86_MACHINE(obj); > > > > + > > > > + x86ms->hyperv_enabled = value; > > > > +} > > > > + > > > > static void x86_machine_initfn(Object *obj) > > > > { > > > > X86MachineState *x86ms = X86_MACHINE(obj); > > > > @@ -1194,6 +1208,16 @@ static void x86_machine_class_init(ObjectClass *oc, void *data) > > > > x86mc->save_tsc_khz = true; > > > > nc->nmi_monitor_handler = x86_nmi; > > > > > > > > + /* Hyper-V features enabled with 'hyperv=on' */ > > > > + x86mc->default_hyperv_features = BIT(HYPERV_FEAT_RELAXED) | > > > > + BIT(HYPERV_FEAT_VAPIC) | BIT(HYPERV_FEAT_TIME) | > > > > + BIT(HYPERV_FEAT_CRASH) | BIT(HYPERV_FEAT_RESET) | > > > > + BIT(HYPERV_FEAT_VPINDEX) | BIT(HYPERV_FEAT_RUNTIME) | > > > > + BIT(HYPERV_FEAT_SYNIC) | BIT(HYPERV_FEAT_STIMER) | > > > > + BIT(HYPERV_FEAT_FREQUENCIES) | BIT(HYPERV_FEAT_REENLIGHTENMENT) | > > > > + BIT(HYPERV_FEAT_TLBFLUSH) | BIT(HYPERV_FEAT_EVMCS) | > > > > + BIT(HYPERV_FEAT_IPI) | BIT(HYPERV_FEAT_STIMER_DIRECT); > > I'd argue that feature bits do not belong to machine code at all. > > If we have to involve machine at all then it should be a set property/value pairs > > that machine will set on CPU object (I'm not convinced that doing it > > from machine code is good idea though). > > The set of default hyperv features needs be defined by the > machine type somehow, we can't avoid that. > > You are correct that the policy could be implemented using > compat_props, but I don't think we should block a patch just > because we're not using a pure QOM property-based interface to > implement that. I'm fine with 1-4/5 patches but not with this one. With this patch I don't agree with inventing special semantics to property handling when it could be done in a typical and consistent way (especially for the sake of convenience). > We need the external interface to be good, though: > > > > [...] > > > > static void x86_cpu_hyperv_realize(X86CPU *cpu) > > > > { > > > > + X86MachineState *x86ms = X86_MACHINE(qdev_get_machine()); > > > > + X86MachineClass *x86mc = X86_MACHINE_GET_CLASS(x86ms); > > > > + uint64_t feat; > > > > size_t len; > > > > > > > > + if (x86ms->hyperv_enabled) { > > > > + feat = x86mc->default_hyperv_features; > > > > + /* Enlightened VMCS is only available on Intel/VMX */ > > > > + if (!cpu_has_vmx(&cpu->env)) { > > > > + feat &= ~BIT(HYPERV_FEAT_EVMCS); > > > > + } > > > > + > > > > + cpu->hyperv_features |= feat; > > that will ignore features user explicitly doesn't want, > > ex: > > -machine hyperv=on -cpu foo,hv-foo=off > > Oops, good point. > > > > > > not sure we would like to introduce such invariant, > > in normal qom property handling the latest set property should have effect > > (all other invariants we have in x86 cpu property semantics are comming from legacy handling > > and I plan to deprecate them (it will affect x86 and sparc cpus) so CPUs will behave like > > any other QOM object when it come to property handling) > > > > anyways it's confusing a bit to have cpu flags to come from 2 different places > > > > -cpu hyperv-use-preset=on,hv-foo=off > > > > looks less confusing and will heave expected effect > > > > > > + } > > > > > > I had to dequeue this because it doesn't compile with > > > CONFIG_USER_ONLY: > > > > > > https://gitlab.com/ehabkost/qemu/-/jobs/916651017 > > > > > > The easiest solution would be to wrap the new code in #ifndef > > > CONFIG_USER_ONLY, but maybe we should try to move all > > > X86Machine-specific code from cpu.c to > > > hw/i386/x86.c:x86_cpu_pre_plug(). > > this looks to me like a preset of feature flags that belongs to CPU, > > and machine code here only as a way to version subset of CPU features. > > > > Is there a way to implement it without modifying machine? > > Maybe there is, but why modifying machine is a problem? 1. it doesn't let do the job properly (realize time is too late) 2. unnecessarily pushes CPU specific logic to machine code, it just doesn't belong there. Sure we can do that here, then some where else and in the end code becomes unmanageable mess. > I agree the interface needs to be clear and consistent, though. > Maybe making it a -cpu option would make this clearer and more > consistent. > > > > > for example versioned CPUs or maybe something like this: > > > > for CLI: > > -cpu hyperv-use-preset=on,hv-foo=off > > In either case, we must clearly define what should happen if the > preset is (HYPERV_FEAT_X | HYPERV_FEAT_Y), and the command line > has: > > -cpu foo,hv-A=on,hv-X=off,hyperv-use-preset=on,hv-B=on,hv-Y=off current x86 cpu code (it doesn't have typical properties handling for keeping legacy semantics), it will basically reorder all features with 'off' value to the end, so hv-X=off will still have an effect. However I plan to deprecate those reordering semantics (x86/sparc cpus), to make it consistent with typical property handling (last set value overwrites any previously set one). That will let us drop custom parsing of -cpu (quite a bit of code) and more importantly make it consistent with -device/device_add cpu-foo. > or: > > -machine hyperv=on -cpu foo,hv-A=on,hv-X=off,hv-B=on,hv-X=off > > Personally, I don't care what the rules are, as long as: 1) they > are clearly defined and documented; 2) they support the use cases > we need to support. I'd like to stick with typical property handling rules, and resort to inventing/using other invariant only if there is no other choice. > An automated test case to make sure we don't break the rules > would be really welcome. > > > > > diff --git a/hw/core/machine.c b/hw/core/machine.c > > index 8d1a90c6cf..8828dcde8e 100644 > > --- a/hw/core/machine.c > > +++ b/hw/core/machine.c > > @@ -35,6 +35,7 @@ GlobalProperty hw_compat_5_0[] = { > > { "vmport", "x-signal-unsupported-cmd", "off" }, > > { "vmport", "x-report-vmx-type", "off" }, > > { "vmport", "x-cmds-v2", "off" }, > > + { "cpu-foo", "hv-preset", "0xXXXX" }, // use compat props to keep old defaults > > + // it will be set before we return from object_new(cpu_type) > > }; > > const size_t hw_compat_5_0_len = G_N_ELEMENTS(hw_compat_5_0); > > > > diff --git a/slirp b/slirp > > --- a/slirp > > +++ b/slirp > > @@ -1 +1 @@ > > -Subproject commit ce94eba2042d52a0ba3d9e252ebce86715e94275 > > +Subproject commit ce94eba2042d52a0ba3d9e252ebce86715e94275-dirty > > diff --git a/target/i386/cpu.c b/target/i386/cpu.c > > index 588f32e136..f0b511ce27 100644 > > --- a/target/i386/cpu.c > > +++ b/target/i386/cpu.c > > @@ -7190,6 +7190,8 @@ static Property x86_cpu_properties[] = { > > > > DEFINE_PROP_UINT32("hv-spinlocks", X86CPU, hyperv_spinlock_attempts, > > HYPERV_SPINLOCK_NEVER_RETRY), > > + DEFINE_PROP_UNIT64("hyperv-preset", X86CPU, hyperv_features_def, 0xYYYYY), > > + // prop_info should define custom setter/getter that will copy hyperv_features_def into hyperv_features > > + // moment "hyperv-use-preset=on" is processed, it will overwrite any previously set > > + // hv-foo but that's fine because user asked for it explictly > > + DEFINE_PROP("hyperv-use-preset", X86CPU, hyperv_use_preset, prop_info, bool), > > We don't need to use custom getters/setters with DEFINE_PROP, if > we can use object_class_property_add_bool(). of cause, I've used DEFINE_PROP just as a possible example. > I dislike custom getters/setters in either case, but maybe we > don't have a choice. Depending on the rules we agree upon above, > custom setters could become avoidable, or they could become a > necessity. I do dislike them too, but sometimes custom setters are convenient as they allow to check if value is valid and let us implement non trivial handling (like in this case) at property setting time. (doing overwites) > > DEFINE_PROP_BIT64("hv-relaxed", X86CPU, hyperv_features, > > HYPERV_FEAT_RELAXED, 0), > > DEFINE_PROP_BIT64("hv-vapic", X86CPU, hyperv_features, >