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=-7.0 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH,MAILING_LIST_MULTI,SIGNED_OFF_BY,SPF_PASS 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 B41C8C43441 for ; Mon, 26 Nov 2018 16:44:33 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 72B4020659 for ; Mon, 26 Nov 2018 16:44:33 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 72B4020659 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=redhat.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-kernel-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726731AbeK0DjK (ORCPT ); Mon, 26 Nov 2018 22:39:10 -0500 Received: from mx1.redhat.com ([209.132.183.28]:57872 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726300AbeK0DjK (ORCPT ); Mon, 26 Nov 2018 22:39:10 -0500 Received: from smtp.corp.redhat.com (int-mx03.intmail.prod.int.phx2.redhat.com [10.5.11.13]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 4433430842AB; Mon, 26 Nov 2018 16:44:31 +0000 (UTC) Received: from [10.36.112.66] (ovpn-112-66.ams2.redhat.com [10.36.112.66]) by smtp.corp.redhat.com (Postfix) with ESMTPS id 82CC4608FA; Mon, 26 Nov 2018 16:44:27 +0000 (UTC) Subject: Re: [PATCH v2 3/4] x86/kvm/hyper-v: direct mode for synthetic timers To: Vitaly Kuznetsov , kvm@vger.kernel.org Cc: =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= , linux-kernel@vger.kernel.org, Roman Kagan , "K. Y. Srinivasan" , Haiyang Zhang , Stephen Hemminger , x86@kernel.org, "Michael Kelley (EOSG)" References: <20181126154732.23025-1-vkuznets@redhat.com> <20181126154732.23025-4-vkuznets@redhat.com> From: Paolo Bonzini Openpgp: preference=signencrypt Autocrypt: addr=pbonzini@redhat.com; prefer-encrypt=mutual; keydata= xsEhBFRCcBIBDqDGsz4K0zZun3jh+U6Z9wNGLKQ0kSFyjN38gMqU1SfP+TUNQepFHb/Gc0E2 CxXPkIBTvYY+ZPkoTh5xF9oS1jqI8iRLzouzF8yXs3QjQIZ2SfuCxSVwlV65jotcjD2FTN04 hVopm9llFijNZpVIOGUTqzM4U55sdsCcZUluWM6x4HSOdw5F5Utxfp1wOjD/v92Lrax0hjiX DResHSt48q+8FrZzY+AUbkUS+Jm34qjswdrgsC5uxeVcLkBgWLmov2kMaMROT0YmFY6A3m1S P/kXmHDXxhe23gKb3dgwxUTpENDBGcfEzrzilWueOeUWiOcWuFOed/C3SyijBx3Av/lbCsHU Vx6pMycNTdzU1BuAroB+Y3mNEuW56Yd44jlInzG2UOwt9XjjdKkJZ1g0P9dwptwLEgTEd3Fo UdhAQyRXGYO8oROiuh+RZ1lXp6AQ4ZjoyH8WLfTLf5g1EKCTc4C1sy1vQSdzIRu3rBIjAvnC tGZADei1IExLqB3uzXKzZ1BZ+Z8hnt2og9hb7H0y8diYfEk2w3R7wEr+Ehk5NQsT2MPI2QBd wEv1/Aj1DgUHZAHzG1QN9S8wNWQ6K9DqHZTBnI1hUlkp22zCSHK/6FwUCuYp1zcAEQEAAc0f UGFvbG8gQm9uemluaSA8Ym9uemluaUBnbnUub3JnPsLBTQQTAQIAIwUCVEJ7AwIbAwcLCQgH AwIBBhUIAgkKCwQWAgMBAh4BAheAAAoJEH4VEAzNNmmxNcwOniaZVLsuy1lW/ntYCA0Caz0i sHpmecK8aWlvL9wpQCk4GlOX9L1emyYXZPmzIYB0IRqmSzAlZxi+A2qm9XOxs5gJ2xqMEXX5 FMtUH3kpkWWJeLqe7z0EoQdUI4EG988uv/tdZyqjUn2XJE+K01x7r3MkUSFz/HZKZiCvYuze VlS0NTYdUt5jBXualvAwNKfxEkrxeHjxgdFHjYWhjflahY7TNRmuqPM/Lx7wAuyoDjlYNE40 Z+Kun4/KjMbjgpcF4Nf3PJQR8qXI6p3so2qsSn91tY7DFSJO6v2HwFJkC2jU95wxfNmTEUZc znXahYbVOwCDJRuPrE5GKFd/XJU9u5hNtr/uYipHij01WXal2cce1S5mn1/HuM1yo1u8xdHy IupCd57EWI948e8BlhpujUCU2tzOb2iYS0kpmJ9/oLVZrOcSZCcCl2P0AaCAsj59z2kwQS9D du0WxUs8waso0Qq6tDEHo8yLCOJDzSz4oojTtWe4zsulVnWV+wu70AioemAT8S6JOtlu60C5 dHgQUD1Tp+ReXpDKXmjbASJx4otvW0qah3o6JaqO79tbDqIvncu3tewwp6c85uZd48JnIOh3 utBAu684nJakbbvZUGikJfxd887ATQRUQnHuAQgAx4dxXO6/Zun0eVYOnr5GRl76+2UrAAem Vv9Yfn2PbDIbxXqLff7oyVJIkw4WdhQIIvvtu5zH24iYjmdfbg8iWpP7NqxUQRUZJEWbx2CR wkMHtOmzQiQ2tSLjKh/cHeyFH68xjeLcinR7jXMrHQK+UCEw6jqi1oeZzGvfmxarUmS0uRuf fAb589AJW50kkQK9VD/9QC2FJISSUDnRC0PawGSZDXhmvITJMdD4TjYrePYhSY4uuIV02v02 8TVAaYbIhxvDY0hUQE4r8ZbGRLn52bEzaIPgl1p/adKfeOUeMReg/CkyzQpmyB1TSk8lDMxQ zCYHXAzwnGi8WU9iuE1P0wARAQABwsEzBBgBAgAJBQJUQnHuAhsMAAoJEH4VEAzNNmmxp1EO oJy0uZggJm7gZKeJ7iUpeX4eqUtqelUw6gU2daz2hE/jsxsTbC/w5piHmk1H1VWDKEM4bQBT uiJ0bfo55SWsUNN+c9hhIX+Y8LEe22izK3w7mRpvGcg+/ZRG4DEMHLP6JVsv5GMpoYwYOmHn plOzCXHvmdlW0i6SrMsBDl9rw4AtIa6bRwWLim1lQ6EM3PWifPrWSUPrPcw4OLSwFk0CPqC4 HYv/7ZnASVkR5EERFF3+6iaaVi5OgBd81F1TCvCX2BEyIDRZLJNvX3TOd5FEN+lIrl26xecz 876SvcOb5SL5SKg9/rCBufdPSjojkGFWGziHiFaYhbuI2E+NfWLJtd+ZvWAAV+O0d8vFFSvr iy9enJ8kxJwhC0ECbSKFY+W1eTIhMD3aeAKY90drozWEyHhENf4l/V+Ja5vOnW+gCDQkGt2Y 1lJAPPSIqZKvHzGShdh8DduC0U3xYkfbGAUvbxeepjgzp0uEnBXfPTy09JGpgWbg0w91GyfT /ujKaGd4vxG2Ei+MMNDmS1SMx7wu0evvQ5kT9NPzyq8R2GIhVSiAd2jioGuTjX6AZCFv3ToO 53DliFMkVTecLptsXaesuUHgL9dKIfvpm+rNXRn9wAwGjk0X/A== Message-ID: Date: Mon, 26 Nov 2018 17:44:24 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.3.0 MIME-Version: 1.0 In-Reply-To: <20181126154732.23025-4-vkuznets@redhat.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit X-Scanned-By: MIMEDefang 2.79 on 10.5.11.13 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.40]); Mon, 26 Nov 2018 16:44:31 +0000 (UTC) Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 26/11/18 16:47, Vitaly Kuznetsov wrote: > Turns out Hyper-V on KVM (as of 2016) will only use synthetic timers > if direct mode is available. With direct mode we notify the guest by > asserting APIC irq instead of sending a SynIC message. > > The implementation uses existing vec_bitmap for letting lapic code > know that we're interested in the particular IRQ's EOI request. We assume > that the same APIC irq won't be used by the guest for both direct mode > stimer and as sint source (especially with AutoEOI semantics). It is > unclear how things should be handled if that's not true. > > Direct mode is also somewhat less expensive; in my testing > stimer_send_msg() takes not less than 1500 cpu cycles and > stimer_notify_direct() can usually be done in 300-400. WS2016 without > Hyper-V, however, always sticks to non-direct version. > > Signed-off-by: Vitaly Kuznetsov > --- > - Changes since v1: avoid open-coding stimer_mark_pending() in > kvm_hv_synic_send_eoi() [Paolo Bonzini] > --- > arch/x86/kvm/hyperv.c | 67 +++++++++++++++++++++++++++++++++++----- > arch/x86/kvm/trace.h | 10 +++--- > arch/x86/kvm/x86.c | 1 + > include/uapi/linux/kvm.h | 1 + > 4 files changed, 67 insertions(+), 12 deletions(-) > > diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c > index eaec15c738df..9533133be566 100644 > --- a/arch/x86/kvm/hyperv.c > +++ b/arch/x86/kvm/hyperv.c > @@ -38,6 +38,9 @@ > > #define KVM_HV_MAX_SPARSE_VCPU_SET_BITS DIV_ROUND_UP(KVM_MAX_VCPUS, 64) > > +static void stimer_mark_pending(struct kvm_vcpu_hv_stimer *stimer, > + bool vcpu_kick); > + > static inline u64 synic_read_sint(struct kvm_vcpu_hv_synic *synic, int sint) > { > return atomic64_read(&synic->sint[sint]); > @@ -53,8 +56,21 @@ static inline int synic_get_sint_vector(u64 sint_value) > static bool synic_has_vector_connected(struct kvm_vcpu_hv_synic *synic, > int vector) > { > + struct kvm_vcpu *vcpu = synic_to_vcpu(synic); > + struct kvm_vcpu_hv *hv_vcpu = vcpu_to_hv_vcpu(vcpu); > + struct kvm_vcpu_hv_stimer *stimer; > int i; > > + for (i = 0; i < ARRAY_SIZE(hv_vcpu->stimer); i++) { > + stimer = &hv_vcpu->stimer[i]; > + if (stimer->config.enable && stimer->config.direct_mode && > + stimer->config.apic_vector == vector) > + return true; > + } > + > + if (vector < HV_SYNIC_FIRST_VALID_VECTOR) > + return false; > + > for (i = 0; i < ARRAY_SIZE(synic->sint); i++) { > if (synic_get_sint_vector(synic_read_sint(synic, i)) == vector) > return true; > @@ -80,14 +96,14 @@ static bool synic_has_vector_auto_eoi(struct kvm_vcpu_hv_synic *synic, > static void synic_update_vector(struct kvm_vcpu_hv_synic *synic, > int vector) > { > - if (vector < HV_SYNIC_FIRST_VALID_VECTOR) > - return; > - > if (synic_has_vector_connected(synic, vector)) > __set_bit(vector, synic->vec_bitmap); > else > __clear_bit(vector, synic->vec_bitmap); > > + if (vector < HV_SYNIC_FIRST_VALID_VECTOR) > + return; > + > if (synic_has_vector_auto_eoi(synic, vector)) > __set_bit(vector, synic->auto_eoi_bitmap); > else > @@ -202,6 +218,7 @@ static void kvm_hv_notify_acked_sint(struct kvm_vcpu *vcpu, u32 sint) > for (idx = 0; idx < ARRAY_SIZE(hv_vcpu->stimer); idx++) { > stimer = &hv_vcpu->stimer[idx]; > if (stimer->msg_pending && stimer->config.enable && > + !stimer->config.direct_mode && > stimer->config.sintx == sint) { > set_bit(stimer->index, > hv_vcpu->stimer_pending_bitmap); > @@ -371,7 +388,9 @@ int kvm_hv_synic_set_irq(struct kvm *kvm, u32 vpidx, u32 sint) > > void kvm_hv_synic_send_eoi(struct kvm_vcpu *vcpu, int vector) > { > + struct kvm_vcpu_hv *hv_vcpu = vcpu_to_hv_vcpu(vcpu); > struct kvm_vcpu_hv_synic *synic = vcpu_to_synic(vcpu); > + struct kvm_vcpu_hv_stimer *stimer; > int i; > > trace_kvm_hv_synic_send_eoi(vcpu->vcpu_id, vector); > @@ -379,6 +398,14 @@ void kvm_hv_synic_send_eoi(struct kvm_vcpu *vcpu, int vector) > for (i = 0; i < ARRAY_SIZE(synic->sint); i++) > if (synic_get_sint_vector(synic_read_sint(synic, i)) == vector) > kvm_hv_notify_acked_sint(vcpu, i); > + > + for (i = 0; i < ARRAY_SIZE(hv_vcpu->stimer); i++) { > + stimer = &hv_vcpu->stimer[i]; > + if (stimer->msg_pending && stimer->config.enable && > + stimer->config.direct_mode && > + stimer->config.apic_vector == vector) > + stimer_mark_pending(stimer, false); > + } > } > > static int kvm_hv_set_sint_gsi(struct kvm *kvm, u32 vpidx, u32 sint, int gsi) > @@ -545,15 +572,25 @@ static int stimer_start(struct kvm_vcpu_hv_stimer *stimer) > static int stimer_set_config(struct kvm_vcpu_hv_stimer *stimer, u64 config, > bool host) > { > - union hv_stimer_config new_config = {.as_uint64 = config}; > + struct kvm_vcpu *vcpu = stimer_to_vcpu(stimer); > + struct kvm_vcpu_hv *hv_vcpu = vcpu_to_hv_vcpu(vcpu); > + union hv_stimer_config new_config = {.as_uint64 = config}, > + old_config = {.as_uint64 = stimer->config.as_uint64}; > > trace_kvm_hv_stimer_set_config(stimer_to_vcpu(stimer)->vcpu_id, > stimer->index, config, host); > > stimer_cleanup(stimer); > - if (stimer->config.enable && new_config.sintx == 0) > + if (old_config.enable && > + !new_config.direct_mode && new_config.sintx == 0) > new_config.enable = 0; > stimer->config.as_uint64 = new_config.as_uint64; > + > + if (old_config.direct_mode) > + synic_update_vector(&hv_vcpu->synic, old_config.apic_vector); > + if (new_config.direct_mode) > + synic_update_vector(&hv_vcpu->synic, new_config.apic_vector); > + > stimer_mark_pending(stimer, false); > return 0; > } > @@ -640,14 +677,28 @@ static int stimer_send_msg(struct kvm_vcpu_hv_stimer *stimer) > stimer->config.sintx, msg); > } > > +static int stimer_notify_direct(struct kvm_vcpu_hv_stimer *stimer) > +{ > + struct kvm_vcpu *vcpu = stimer_to_vcpu(stimer); > + struct kvm_lapic_irq irq = { > + .delivery_mode = APIC_DM_FIXED, > + .vector = stimer->config.apic_vector > + }; > + > + return !kvm_apic_set_irq(vcpu, &irq, NULL); > +} > + > static void stimer_expiration(struct kvm_vcpu_hv_stimer *stimer) > { > - int r; > + int r, direct = stimer->config.direct_mode; > > stimer->msg_pending = true; > - r = stimer_send_msg(stimer); > + if (!direct) > + r = stimer_send_msg(stimer); > + else > + r = stimer_notify_direct(stimer); > trace_kvm_hv_stimer_expiration(stimer_to_vcpu(stimer)->vcpu_id, > - stimer->index, r); > + stimer->index, direct, r); > if (!r) { > stimer->msg_pending = false; > if (!(stimer->config.periodic)) > diff --git a/arch/x86/kvm/trace.h b/arch/x86/kvm/trace.h > index 0659465a745c..705f40ae2532 100644 > --- a/arch/x86/kvm/trace.h > +++ b/arch/x86/kvm/trace.h > @@ -1254,24 +1254,26 @@ TRACE_EVENT(kvm_hv_stimer_callback, > * Tracepoint for stimer_expiration. > */ > TRACE_EVENT(kvm_hv_stimer_expiration, > - TP_PROTO(int vcpu_id, int timer_index, int msg_send_result), > - TP_ARGS(vcpu_id, timer_index, msg_send_result), > + TP_PROTO(int vcpu_id, int timer_index, int direct, int msg_send_result), > + TP_ARGS(vcpu_id, timer_index, direct, msg_send_result), > > TP_STRUCT__entry( > __field(int, vcpu_id) > __field(int, timer_index) > + __field(int, direct) > __field(int, msg_send_result) > ), > > TP_fast_assign( > __entry->vcpu_id = vcpu_id; > __entry->timer_index = timer_index; > + __entry->direct = direct; > __entry->msg_send_result = msg_send_result; > ), > > - TP_printk("vcpu_id %d timer %d msg send result %d", > + TP_printk("vcpu_id %d timer %d direct %d send result %d", > __entry->vcpu_id, __entry->timer_index, > - __entry->msg_send_result) > + __entry->direct, __entry->msg_send_result) > ); > > /* > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > index 5cd5647120f2..b21b5ceb8d26 100644 > --- a/arch/x86/kvm/x86.c > +++ b/arch/x86/kvm/x86.c > @@ -2997,6 +2997,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext) > case KVM_CAP_HYPERV_TLBFLUSH: > case KVM_CAP_HYPERV_SEND_IPI: > case KVM_CAP_HYPERV_ENLIGHTENED_VMCS: > + case KVM_CAP_HYPERV_STIMER_DIRECT: > case KVM_CAP_PCI_SEGMENT: > case KVM_CAP_DEBUGREGS: > case KVM_CAP_X86_ROBUST_SINGLESTEP: > diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h > index 2b7a652c9fa4..b8da14cee8e5 100644 > --- a/include/uapi/linux/kvm.h > +++ b/include/uapi/linux/kvm.h > @@ -975,6 +975,7 @@ struct kvm_ppc_resize_hpt { > #define KVM_CAP_HYPERV_ENLIGHTENED_VMCS 163 > #define KVM_CAP_EXCEPTION_PAYLOAD 164 > #define KVM_CAP_ARM_VM_IPA_SIZE 165 > +#define KVM_CAP_HYPERV_STIMER_DIRECT 166 I wonder if all these capabilities shouldn't be replaced by a single KVM_GET_HYPERV_SUPPORTED_CPUID ioctl, or something like that. If you can do it for 4.21, before this one cap is crystallized into userspace API, that would be great. :) Paolo