Received: by 2002:a05:7412:8d10:b0:f3:1519:9f41 with SMTP id bj16csp5902647rdb; Thu, 14 Dec 2023 03:05:11 -0800 (PST) X-Google-Smtp-Source: AGHT+IF1mQMmwatmmyYInbju+JD2UqDyb4As9RRB7DIxiyjkd4eZLYhv0bnDJBT1S9NzzkD+SPfY X-Received: by 2002:a17:90a:34c2:b0:28b:812:fe96 with SMTP id m2-20020a17090a34c200b0028b0812fe96mr160088pjf.62.1702551911380; Thu, 14 Dec 2023 03:05:11 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1702551911; cv=none; d=google.com; s=arc-20160816; b=FdlMsGfvb+rx153+vkjzSSdOYizL+Ss8uAYiv1juAB9MYXtsx513N/lXimz/8y0nB5 yaI3I79CeMsrT60KvbF0libgDIQVdQ5ZSjhpLctpYI9lnCoJmRDmwGnY5bdXc8JxdIh4 GjVWZuuuuAfpZmOqyc1yG99EsaNnlJDtiXirPvSmz76fc6z7jvXGKAX1gC2vgLT1IRs5 J2gf/nwM1m5BU9tTv3JJqluvROFzIht1dS3ADcREQwyPivIeUtxZD3VndmhwQobU6/0i UBTU9bil9NMeXlT+SyqhMJ1CJY5ZrYpAb13RK4FywSxNt1gk+czrSmtKn1dMlwrpHAd4 mPYg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:message-id:date:subject:cc:to:from :dkim-signature; bh=yWP+2D1eULPusEz0RCI9Wdbqiu1t2gWF7XI5hhSVxBk=; fh=flttZmas51TMfnCnDh2IPQNTIdR4x+7WxTSos7vPkVw=; b=jcKv6kP0zfLcIr/WxBPTizvXX0kat55EzbXJQpKuq4dc2JNEM5WYg5gPRYme9PHkng hW7B+7m9LnLqM9bgEDhD4YANyWHUSjkBv6gYB1f1CsQrv45VFNW/Qh2idxovvjZnSZNK 6PCFGRHHRRo+lETmV26eBdR1ISpW8hB+s+LznM3Edfx53WajZf7O6AnXu/v3SWOg8sgm kJVk3ZHln9HhbCMxql5AKTyZdv8hKC2+ItSMUE9fV0mCgXXNL2MDFEXBgypECki3coTK oh/L8v50TXzqHDiu/qXENps1+uFSzHB5Jw/f4qEQ/9tRhOX7UE2nfbmiDyF53O2SfYkd rSRA== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@intel.com header.s=Intel header.b=P7DrjrDl; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:3 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=intel.com Return-Path: Received: from lipwig.vger.email (lipwig.vger.email. [2620:137:e000::3:3]) by mx.google.com with ESMTPS id g20-20020a17090ace9400b0028aecb0d2b3si2092891pju.15.2023.12.14.03.05.10 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 14 Dec 2023 03:05:11 -0800 (PST) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:3 as permitted sender) client-ip=2620:137:e000::3:3; Authentication-Results: mx.google.com; dkim=pass header.i=@intel.com header.s=Intel header.b=P7DrjrDl; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:3 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: from out1.vger.email (depot.vger.email [IPv6:2620:137:e000::3:0]) by lipwig.vger.email (Postfix) with ESMTP id 9D19D8145940; Thu, 14 Dec 2023 03:05:07 -0800 (PST) X-Virus-Status: Clean X-Virus-Scanned: clamav-milter 0.103.11 at lipwig.vger.email Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1443895AbjLNLEt (ORCPT + 99 others); Thu, 14 Dec 2023 06:04:49 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:37656 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1443893AbjLNLEm (ORCPT ); Thu, 14 Dec 2023 06:04:42 -0500 Received: from mgamail.intel.com (mgamail.intel.com [192.198.163.8]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 75B2C11A; Thu, 14 Dec 2023 03:04:42 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1702551883; x=1734087883; h=from:to:cc:subject:date:message-id; bh=PPJ0qo+v4gU51NZ6QR1Rk2xhM3fO+C3owLqS4ypDdH8=; b=P7DrjrDlccdgzyGxKAYA6wtrd7xo1xXYGpsL2ZgeWrA23jwC5KFa2Zi4 +AXaZSXzbtC4QQW7h9ZSP0DkjmweU9vhDacREMtB54Re9Kgv0uFXYCOHS S9Rez4M5HH8fYiNYjsjjF7qgRk2yfbrqy3FbPe+2C7HsKPb1jaJAnlWvl yuZFMki2/s3wxxIQlIs3RnHp022jec6b8OTO2tfB2R9SwEq+nnI4T8v1Z tOPCCcVOq5AirnrK9a5ajO0LOSO2qlWu2Byew1weV0HI88pto3FA9v6Ip TxGZqoKHvX2QiA0SIL3jIDSIooTTHJi76wrODzggkVrHcMs4Vh8cCZRgz Q==; X-IronPort-AV: E=McAfee;i="6600,9927,10923"; a="8500543" X-IronPort-AV: E=Sophos;i="6.04,275,1695711600"; d="scan'208";a="8500543" Received: from orviesa001.jf.intel.com ([10.64.159.141]) by fmvoesa102.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 14 Dec 2023 03:04:42 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.04,275,1695711600"; d="scan'208";a="22358938" Received: from yzhao56-desk.sh.intel.com ([10.239.159.62]) by smtpauth.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 14 Dec 2023 03:04:37 -0800 From: Yan Zhao To: kvm@vger.kernel.org, linux-kernel@vger.kernel.org, dri-devel@lists.freedesktop.org Cc: pbonzini@redhat.com, seanjc@google.com, olvaffe@gmail.com, kevin.tian@intel.com, zhiyuan.lv@intel.com, zhenyu.z.wang@intel.com, yongwei.ma@intel.com, vkuznets@redhat.com, wanpengli@tencent.com, jmattson@google.com, joro@8bytes.org, gurchetansingh@chromium.org, kraxel@redhat.com, Yan Zhao Subject: [RFC PATCH] KVM: Introduce KVM VIRTIO device Date: Thu, 14 Dec 2023 18:35:20 +0800 Message-Id: <20231214103520.7198-1-yan.y.zhao@intel.com> X-Mailer: git-send-email 2.17.1 X-Spam-Status: No, score=-0.9 required=5.0 tests=DKIMWL_WL_HIGH,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,T_SCC_BODY_TEXT_LINE autolearn=unavailable autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lipwig.vger.email Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-Greylist: Sender passed SPF test, not delayed by milter-greylist-4.6.4 (lipwig.vger.email [0.0.0.0]); Thu, 14 Dec 2023 03:05:07 -0800 (PST) Introduce a new KVM device that represents a virtio device in order to allow user to indicate device hardware's noncoherent DMA status. Motivation === A virtio GPU device may want to configure GPU hardware to work in noncoherent mode, i.e. some of its DMAs do not snoop CPU caches. This is generally for performance consideration. In certain platform, GFX performance can improve 20+% with DMAs going to noncoherent path. This noncoherent DMA mode works in below sequence: 1. Host backend driver programs hardware not to snoop memory of target DMA buffer. 2. Host backend driver indicates guest frontend driver to program guest PAT to WC for target DMA buffer. 3. Guest frontend driver writes to the DMA buffer without clflush stuffs. 4. Hardware does noncoherent DMA to the target buffer. In this noncoherent DMA mode, both guest and hardware regard a DMA buffer as not cached. So, if KVM forces the effective memory type of this DMA buffer to be WB, hardware DMA may read incorrect data and cause misc failures. Therefore, we introduce a new KVM virtio device to let KVM be aware that the virtio device hardware is working in noncoherent mode. Then, KVM will honor guest PAT type in Intel's platform. For a virtio device model (e.g. QEMU), if it knows device hardware is to be configured to work in a. noncoherent mode, - on device realization, it can create a KVM virtio device and set device attr to increase KVM noncoherent DMA count; - on device unrealization, destroy the KVM virtio device to decrease KVM noncoherent DMA count. b. coherent mode, just do nothing. Security === The biggest concern for KVM to honor guest's memory type in Intel platform is page aliasing issue. - For host MMIO, it's not a concern as KVM VMX programs EPT memory type to UC (which will overwrite all guest PAT type except WC) no matter guest memory type is honored or not. - For host non-MMIO pages, * virtio guest frontend and host backend driver should be synced to use the same memory type to map a buffer. Otherwise, there will be potential problem for incorrect memory data. But this will only impact the buggy guest alone. * for live migration, as QEMU will read all guest memory during live migration, page aliasing could happen. Current thinking is to disable live migration if a virtio device has indicated its noncoherent state. As a follow-up, we can discuss other solutions. e.g. (a) switching back to coherent path before starting live migration. (b) read/write of guest memory with clflush during live migration. Implementation Consideration === There is a previous series [1] from google to serve the same purpose to let KVM be aware of virtio GPU's noncoherent DMA status. That series requires a new memslot flag, and special memslots in user space. We don't choose to use memslot flag to request honoring guest memory type. Instead we hope to make the honoring request to be explicit (not tied to a memslot flag). This is because once guest memory type is honored, not only memory used by guest virtio device, but all guest memory is facing page aliasing issue potentially. KVM needs a generic solution to take care of page aliasing issue rather than counting on memory type of a special memslot being aligned in host and guest. (we can discuss what a generic solution to handle page aliasing issue will look like in later follow-up series). On the other hand, we choose to introduce a KVM virtio device rather than just provide an ioctl to wrap kvm_arch_[un]register_noncoherent_dma() directly, which is based on considerations that 1. Explicitly limit the "register noncoherent DMA" ability to virtio devices. 2. Provide a better encapsulation. Repeated setting noncoherent attribute to a KVM virtio device will only increase KVM noncoherent DMA count for once. KVM noncohrent DMA count will be decreased automatically when KVM virtio device is closed. 3. The KVM virtio device can be extended to let KVM know more info about the device to introduce non-coherent DMA. Example QEMU usage === - on virtio device realize: struct kvm_create_device cd = { .type = KVM_DEV_TYPE_VIRTIO, }; kvm_vm_ioctl(kvm_state, KVM_CREATE_DEVICE, &cd); struct kvm_device_attr attr = { .group = KVM_DEV_VIRTIO_NONCOHERENT, .attr = KVM_DEV_VIRTIO_NONCOHERENT_SET, }; ioctl(cd.fd, KVM_SET_DEVICE_ATTR, &attr); g->kvm_device_fd = cd.fd; - on virtio device unrealize close(g->kvm_device_fd); Reviewed-by: Zhenyu Wang Link: https://patchwork.kernel.org/project/dri-devel/cover/20200213213036.207625-1-olvaffe@gmail.com/ [1] Signed-off-by: Yan Zhao --- arch/x86/kvm/Kconfig | 1 + include/uapi/linux/kvm.h | 5 ++ virt/kvm/Kconfig | 3 + virt/kvm/Makefile.kvm | 1 + virt/kvm/kvm_main.c | 8 +++ virt/kvm/virtio.c | 121 +++++++++++++++++++++++++++++++++++++++ virt/kvm/virtio.h | 18 ++++++ 7 files changed, 157 insertions(+) create mode 100644 virt/kvm/virtio.c create mode 100644 virt/kvm/virtio.h diff --git a/arch/x86/kvm/Kconfig b/arch/x86/kvm/Kconfig index b07247b0b958..9f7223d298a2 100644 --- a/arch/x86/kvm/Kconfig +++ b/arch/x86/kvm/Kconfig @@ -44,6 +44,7 @@ config KVM select KVM_XFER_TO_GUEST_WORK select KVM_GENERIC_DIRTYLOG_READ_PROTECT select KVM_VFIO + select KVM_VIRTIO select INTERVAL_TREE select HAVE_KVM_PM_NOTIFIER if PM select KVM_GENERIC_HARDWARE_ENABLING diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h index b1f92a0edc35..7f57fa902a0a 100644 --- a/include/uapi/linux/kvm.h +++ b/include/uapi/linux/kvm.h @@ -1394,6 +1394,9 @@ struct kvm_device_attr { #define KVM_DEV_VFIO_GROUP_DEL KVM_DEV_VFIO_FILE_DEL #define KVM_DEV_VFIO_GROUP_SET_SPAPR_TCE 3 +#define KVM_DEV_VIRTIO_NONCOHERENT 1 +#define KVM_DEV_VIRTIO_NONCOHERENT_SET 1 + enum kvm_device_type { KVM_DEV_TYPE_FSL_MPIC_20 = 1, #define KVM_DEV_TYPE_FSL_MPIC_20 KVM_DEV_TYPE_FSL_MPIC_20 @@ -1417,6 +1420,8 @@ enum kvm_device_type { #define KVM_DEV_TYPE_ARM_PV_TIME KVM_DEV_TYPE_ARM_PV_TIME KVM_DEV_TYPE_RISCV_AIA, #define KVM_DEV_TYPE_RISCV_AIA KVM_DEV_TYPE_RISCV_AIA + KVM_DEV_TYPE_VIRTIO, +#define KVM_DEV_TYPE_VIRTIO KVM_DEV_TYPE_VIRTIO KVM_DEV_TYPE_MAX, }; diff --git a/virt/kvm/Kconfig b/virt/kvm/Kconfig index 6793211a0b64..5dcba034593c 100644 --- a/virt/kvm/Kconfig +++ b/virt/kvm/Kconfig @@ -56,6 +56,9 @@ config HAVE_KVM_CPU_RELAX_INTERCEPT config KVM_VFIO bool +config KVM_VIRTIO + bool + config HAVE_KVM_INVALID_WAKEUPS bool diff --git a/virt/kvm/Makefile.kvm b/virt/kvm/Makefile.kvm index 724c89af78af..614c4c4fe50e 100644 --- a/virt/kvm/Makefile.kvm +++ b/virt/kvm/Makefile.kvm @@ -7,6 +7,7 @@ KVM ?= ../../../virt/kvm kvm-y := $(KVM)/kvm_main.o $(KVM)/eventfd.o $(KVM)/binary_stats.o kvm-$(CONFIG_KVM_VFIO) += $(KVM)/vfio.o +kvm-$(CONFIG_KVM_VIRTIO) += $(KVM)/virtio.o kvm-$(CONFIG_KVM_MMIO) += $(KVM)/coalesced_mmio.o kvm-$(CONFIG_KVM_ASYNC_PF) += $(KVM)/async_pf.o kvm-$(CONFIG_HAVE_KVM_IRQ_ROUTING) += $(KVM)/irqchip.o diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index acd67fb40183..dca2040368da 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -61,6 +61,7 @@ #include "async_pf.h" #include "kvm_mm.h" #include "vfio.h" +#include "virtio.h" #include @@ -6453,6 +6454,10 @@ int kvm_init(unsigned vcpu_size, unsigned vcpu_align, struct module *module) if (WARN_ON_ONCE(r)) goto err_vfio; + r = kvm_virtio_ops_init(); + if (WARN_ON_ONCE(r)) + goto err_virtio; + kvm_gmem_init(module); /* @@ -6468,6 +6473,8 @@ int kvm_init(unsigned vcpu_size, unsigned vcpu_align, struct module *module) return 0; err_register: + kvm_virtio_ops_exit(); +err_virtio: kvm_vfio_ops_exit(); err_vfio: kvm_async_pf_deinit(); @@ -6503,6 +6510,7 @@ void kvm_exit(void) free_cpumask_var(per_cpu(cpu_kick_mask, cpu)); kmem_cache_destroy(kvm_vcpu_cache); kvm_vfio_ops_exit(); + kvm_virtio_ops_exit(); kvm_async_pf_deinit(); #ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING unregister_syscore_ops(&kvm_syscore_ops); diff --git a/virt/kvm/virtio.c b/virt/kvm/virtio.c new file mode 100644 index 000000000000..dbb25d9784c5 --- /dev/null +++ b/virt/kvm/virtio.c @@ -0,0 +1,121 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * KVM-VIRTIO device + * + */ +#include +#include +#include +#include +#include "virtio.h" + +struct kvm_virtio { + struct mutex lock; + bool noncoherent; +}; + +static int kvm_virtio_set_noncoherent(struct kvm_device *dev, long attr, + void __user *arg) +{ + struct kvm_virtio *kv = dev->private; + + /* + * Currently, only set to noncoherent is allowed, and therefore a virtio + * device is not allowed to switch back to coherent once it's set to + * noncoherent. + * User arg is also not checked as the attr name has indicated that the + * purpose is to set to noncoherent. + */ + if (attr != KVM_DEV_VIRTIO_NONCOHERENT_SET) + return -ENXIO; + + mutex_lock(&kv->lock); + if (kv->noncoherent) + goto out; + + kv->noncoherent = true; + kvm_arch_register_noncoherent_dma(dev->kvm); +out: + mutex_unlock(&kv->lock); + return 0; +} + +static int kvm_virtio_set_attr(struct kvm_device *dev, + struct kvm_device_attr *attr) +{ + switch (attr->group) { + case KVM_DEV_VIRTIO_NONCOHERENT: + return kvm_virtio_set_noncoherent(dev, attr->attr, + u64_to_user_ptr(attr->addr)); + } + + return -ENXIO; +} + +static int kvm_virtio_has_attr(struct kvm_device *dev, + struct kvm_device_attr *attr) +{ + switch (attr->group) { + case KVM_DEV_VIRTIO_NONCOHERENT: + switch (attr->attr) { + case KVM_DEV_VIRTIO_NONCOHERENT_SET: + return 0; + } + + break; + } + + return -ENXIO; +} + +static void kvm_virtio_release(struct kvm_device *dev) +{ + struct kvm_virtio *kv = dev->private; + + if (kv->noncoherent) + kvm_arch_unregister_noncoherent_dma(dev->kvm); + kfree(kv); + kfree(dev); /* alloc by kvm_ioctl_create_device, free by .release */ +} + +static int kvm_virtio_create(struct kvm_device *dev, u32 type); + +static struct kvm_device_ops kvm_virtio_ops = { + .name = "kvm-virtio", + .create = kvm_virtio_create, + .release = kvm_virtio_release, + .set_attr = kvm_virtio_set_attr, + .has_attr = kvm_virtio_has_attr, +}; + +static int kvm_virtio_create(struct kvm_device *dev, u32 type) +{ + struct kvm_virtio *kv; + + if (type != KVM_DEV_TYPE_VIRTIO) + return -ENODEV; + + /* + * This kvm_virtio device is created per virtio device. + * Its default noncoherent state is false. + */ + kv = kzalloc(sizeof(*kv), GFP_KERNEL_ACCOUNT); + if (!kv) + return -ENOMEM; + + mutex_init(&kv->lock); + + dev->private = kv; + + return 0; +} + +int kvm_virtio_ops_init(void) +{ + return kvm_register_device_ops(&kvm_virtio_ops, KVM_DEV_TYPE_VIRTIO); +} + +void kvm_virtio_ops_exit(void) +{ + kvm_unregister_device_ops(KVM_DEV_TYPE_VIRTIO); +} diff --git a/virt/kvm/virtio.h b/virt/kvm/virtio.h new file mode 100644 index 000000000000..0353398ee1f2 --- /dev/null +++ b/virt/kvm/virtio.h @@ -0,0 +1,18 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef __KVM_VIRTIO_H +#define __KVM_VIRTIO_H + +#ifdef CONFIG_KVM_VIRTIO +int kvm_virtio_ops_init(void); +void kvm_virtio_ops_exit(void); +#else +static inline int kvm_virtio_ops_init(void) +{ + return 0; +} +static inline void kvm_virtio_ops_exit(void) +{ +} +#endif + +#endif base-commit: 8ed26ab8d59111c2f7b86d200d1eb97d2a458fd1 -- 2.17.1