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=-8.9 required=3.0 tests=DKIMWL_WL_HIGH,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,MAILING_LIST_MULTI, MENTIONS_GIT_HOSTING,SPF_HELO_NONE,SPF_PASS,USER_AGENT_GIT autolearn=unavailable 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 3A972C10F29 for ; Mon, 9 Mar 2020 21:44:39 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 0CAD024649 for ; Mon, 9 Mar 2020 21:44:39 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (1024-bit key) header.d=redhat.com header.i=@redhat.com header.b="KfggPge5" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1727064AbgCIVod (ORCPT ); Mon, 9 Mar 2020 17:44:33 -0400 Received: from us-smtp-delivery-1.mimecast.com ([205.139.110.120]:30475 "EHLO us-smtp-1.mimecast.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1726439AbgCIVod (ORCPT ); Mon, 9 Mar 2020 17:44:33 -0400 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1583790271; 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; bh=+ICtR2v0xNomH1WLkex46Nm+YSPs83aCqTW2zuVdMA4=; b=KfggPge5R51zEnrozMLwxsTSWUAjVuCTznd3SN1LJhpj8GOs8BtmvloEFdB4wlXOasRvCZ 0CFcakKCEkA8Yu9xMNlfzd3o/YoyMTLlZafN53U9S9xBddwjHWNVogRt00dKC+eUrbDC0b jkU29E7vuTNPHqEswmktUx8X/2dsHeQ= Received: from mail-qv1-f70.google.com (mail-qv1-f70.google.com [209.85.219.70]) (Using TLS) by relay.mimecast.com with ESMTP id us-mta-28-1U5yz-hBN1Coeo3Cx-Tv6g-1; Mon, 09 Mar 2020 17:44:28 -0400 X-MC-Unique: 1U5yz-hBN1Coeo3Cx-Tv6g-1 Received: by mail-qv1-f70.google.com with SMTP id v19so1570309qvk.9 for ; Mon, 09 Mar 2020 14:44:27 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:mime-version :content-transfer-encoding; bh=+ICtR2v0xNomH1WLkex46Nm+YSPs83aCqTW2zuVdMA4=; b=DbjFGwVBPZdxF5S1W2nktxPFm99wA9M7hg6ksW+9I5hLWcDV+wlQzwdNtpXpWpNY1s 8gcyp7Td35pe86+SBIVw8yMXLq9qJ9e/iAC04HrlnoPslBLW540c1focSSesxY4OS+E5 gPbaTDrHKCm1da4POTh7XyKbFC1YluYIafxYwLJYEMbS98tPyvVM78cJfrYimrMiLi9u 6gAmaR6ZxOGfphO4SZAXSZVpNnM4yRAM3uA2/ikCRe6qvVT9ZbgUaLDaLCu7onoUgg4k HCJ2rrJJo9Ops9io4qk/akqcF4EnI+iN9DcAq0Tlm9ER9OdGWBPrNhOp0yYEcfw6kgR3 SRrw== X-Gm-Message-State: ANhLgQ0DN+TugVv3mSlEKPOinfpd1gcRBgPPzsxMcTL7g8b6gdvrnrfM Sz14xbqyR1E4Qp/5ZoqO5CRR3bFCG+JmzYacqYGRcDS0Bt9UrHc5gm2cIsxKzmgSFSfEe+J1gcv cwZy2f21iyP8phbx/fksnKzZH X-Received: by 2002:a37:4ec1:: with SMTP id c184mr2019109qkb.0.1583790266997; Mon, 09 Mar 2020 14:44:26 -0700 (PDT) X-Google-Smtp-Source: ADFU+vstAS1kAHk1yzYK0Qv2/bUTpVaVnkqEMcxpVeMGOT8PFcgZlWN3qpoiuGO6oI0meC5+JHYWaw== X-Received: by 2002:a37:4ec1:: with SMTP id c184mr2019077qkb.0.1583790266685; Mon, 09 Mar 2020 14:44:26 -0700 (PDT) Received: from xz-x1.redhat.com ([2607:9880:19c0:32::2]) by smtp.gmail.com with ESMTPSA id 82sm10504502qko.91.2020.03.09.14.44.25 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 09 Mar 2020 14:44:25 -0700 (PDT) From: Peter Xu To: linux-kernel@vger.kernel.org, kvm@vger.kernel.org Cc: Yan Zhao , Jason Wang , Alex Williamson , peterx@redhat.com, Vitaly Kuznetsov , "Dr . David Alan Gilbert" , Christophe de Dinechin , "Michael S . Tsirkin" , Kevin Tian , Paolo Bonzini , Sean Christopherson Subject: [PATCH v6 00/14] KVM: Dirty ring interface Date: Mon, 9 Mar 2020 17:44:10 -0400 Message-Id: <20200309214424.330363-1-peterx@redhat.com> X-Mailer: git-send-email 2.24.1 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org KVM branch:=0D https://github.com/xzpeter/linux/tree/kvm-dirty-ring=0D =0D QEMU branch for testing:=0D https://github.com/xzpeter/qemu/tree/kvm-dirty-ring=0D =0D v6:=0D - fix typo in test case (LOG_MODE_CLERA_LOG) [Dave]=0D - fix s390 build (again), by moving kvm_reset_dirty_gfn into=0D kvm_dirty_ring.c, with some macro movements [syzbot]=0D =0D For previous versions, please refer to:=0D =0D V1: https://lore.kernel.org/kvm/20191129213505.18472-1-peterx@redhat.com=0D V2: https://lore.kernel.org/kvm/20191221014938.58831-1-peterx@redhat.com=0D V3: https://lore.kernel.org/kvm/20200109145729.32898-1-peterx@redhat.com=0D V4: https://lore.kernel.org/kvm/20200205025105.367213-1-peterx@redhat.com=0D V5: https://lore.kernel.org/kvm/20200304174947.69595-1-peterx@redhat.com=0D =0D Overview=0D =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=0D =0D This is a continued work from Lei Cao and Paolo=0D Bonzini on the KVM dirty ring interface.=0D =0D The new dirty ring interface is another way to collect dirty pages for=0D the virtual machines. It is different from the existing dirty logging=0D interface in a few ways, majorly:=0D =0D - Data format: The dirty data was in a ring format rather than a=0D bitmap format, so dirty bits to sync for dirty logging does not=0D depend on the size of guest memory any more, but speed of=0D dirtying. Also, the dirty ring is per-vcpu, while the dirty=0D bitmap is per-vm.=0D =0D - Data copy: The sync of dirty pages does not need data copy any more,=0D but instead the ring is shared between the userspace and kernel by=0D page sharings (mmap() on vcpu fd)=0D =0D - Interface: Instead of using the old KVM_GET_DIRTY_LOG,=0D KVM_CLEAR_DIRTY_LOG interfaces, the new ring uses the new=0D KVM_RESET_DIRTY_RINGS ioctl when we want to reset the collected=0D dirty pages to protected mode again (works like=0D KVM_CLEAR_DIRTY_LOG, but ring based). To collecting dirty bits,=0D we only need to read the ring data, no ioctl is needed.=0D =0D Ring Layout=0D =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=0D =0D KVM dirty ring is per-vcpu. Each ring is an array of kvm_dirty_gfn=0D defined as:=0D =0D struct kvm_dirty_gfn {=0D __u32 flags;=0D __u32 slot; /* as_id | slot_id */=0D __u64 offset;=0D };=0D =0D Each GFN is a state machine itself. The state is embeded in the flags=0D field, as defined in the uapi header:=0D =0D /*=0D * KVM dirty GFN flags, defined as:=0D *=0D * |---------------+---------------+--------------|=0D * | bit 1 (reset) | bit 0 (dirty) | Status |=0D * |---------------+---------------+--------------|=0D * | 0 | 0 | Invalid GFN |=0D * | 0 | 1 | Dirty GFN |=0D * | 1 | X | GFN to reset |=0D * |---------------+---------------+--------------|=0D *=0D * Lifecycle of a dirty GFN goes like:=0D *=0D * dirtied collected reset=0D * 00 -----------> 01 -------------> 1X -------+=0D * ^ |=0D * | |=0D * +------------------------------------------+=0D *=0D * The userspace program is only responsible for the 01->1X state=0D * conversion (to collect dirty bits). Also, it must not skip any=0D * dirty bits so that dirty bits are always collected in sequence.=0D */=0D =0D Testing=0D =3D=3D=3D=3D=3D=3D=3D=0D =0D This series provided both the implementation of the KVM dirty ring and=0D the test case. Also I've implemented the QEMU counterpart that can=0D run with the new KVM, link can be found at the top of the cover=0D letter. However that's still a very initial version which is prone to=0D change and future optimizations.=0D =0D I did some measurement with the new method with 24G guest running some=0D dirty workload, I don't see any speedup so far, even in some heavy=0D dirty load it'll be slower (e.g., when 800MB/s random dirty rate, kvm=0D dirty ring takes average of ~73s to complete migration while dirty=0D logging only needs average of ~55s). However that's understandable=0D because 24G guest means only 1M dirty bitmap, that's still a suitable=0D case for dirty logging. Meanwhile heavier workload means worst case=0D for dirty ring.=0D =0D More tests are welcomed if there's bigger host/guest, especially on=0D COLO-like workload.=0D =0D Please review, thanks.=0D =0D Peter Xu (14):=0D KVM: X86: Change parameter for fast_page_fault tracepoint=0D KVM: Cache as_id in kvm_memory_slot=0D KVM: X86: Don't track dirty for KVM_SET_[TSS_ADDR|IDENTITY_MAP_ADDR]=0D KVM: Pass in kvm pointer into mark_page_dirty_in_slot()=0D KVM: X86: Implement ring-based dirty memory tracking=0D KVM: Make dirty ring exclusive to dirty bitmap log=0D KVM: Don't allocate dirty bitmap if dirty ring is enabled=0D KVM: selftests: Always clear dirty bitmap after iteration=0D KVM: selftests: Sync uapi/linux/kvm.h to tools/=0D KVM: selftests: Use a single binary for dirty/clear log test=0D KVM: selftests: Introduce after_vcpu_run hook for dirty log test=0D KVM: selftests: Add dirty ring buffer test=0D KVM: selftests: Let dirty_log_test async for dirty ring test=0D KVM: selftests: Add "-c" parameter to dirty log test=0D =0D Documentation/virt/kvm/api.rst | 123 +++++=0D arch/x86/include/asm/kvm_host.h | 6 +-=0D arch/x86/include/uapi/asm/kvm.h | 1 +=0D arch/x86/kvm/Makefile | 3 +-=0D arch/x86/kvm/mmu/mmu.c | 10 +-=0D arch/x86/kvm/mmutrace.h | 9 +-=0D arch/x86/kvm/svm.c | 9 +-=0D arch/x86/kvm/vmx/vmx.c | 85 +--=0D arch/x86/kvm/x86.c | 49 +-=0D include/linux/kvm_dirty_ring.h | 103 ++++=0D include/linux/kvm_host.h | 19 +=0D include/trace/events/kvm.h | 78 +++=0D include/uapi/linux/kvm.h | 53 ++=0D tools/include/uapi/linux/kvm.h | 44 ++=0D tools/testing/selftests/kvm/Makefile | 2 -=0D .../selftests/kvm/clear_dirty_log_test.c | 2 -=0D tools/testing/selftests/kvm/dirty_log_test.c | 488 ++++++++++++++++--=0D .../testing/selftests/kvm/include/kvm_util.h | 4 +=0D tools/testing/selftests/kvm/lib/kvm_util.c | 67 +++=0D .../selftests/kvm/lib/kvm_util_internal.h | 4 +=0D virt/kvm/dirty_ring.c | 195 +++++++=0D virt/kvm/kvm_main.c | 164 +++++-=0D 22 files changed, 1393 insertions(+), 125 deletions(-)=0D create mode 100644 include/linux/kvm_dirty_ring.h=0D delete mode 100644 tools/testing/selftests/kvm/clear_dirty_log_test.c=0D create mode 100644 virt/kvm/dirty_ring.c=0D =0D -- =0D 2.24.1=0D =0D