From mboxrd@z Thu Jan 1 00:00:00 1970 From: Laurent Vivier Subject: Re: [PATCH] KVM: Device Assignment with VT-d Date: Tue, 16 Sep 2008 14:42:15 +0200 Message-ID: <1221568935.4172.22.camel@frecb07144> References: <20080916005853.7876D250DB2@il.qumranet.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: kvm@vger.kernel.org To: Avi Kivity Return-path: Received: from ecfrec.frec.bull.fr ([129.183.4.8]:36464 "EHLO ecfrec.frec.bull.fr" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750864AbYIPMmV (ORCPT ); Tue, 16 Sep 2008 08:42:21 -0400 In-Reply-To: <20080916005853.7876D250DB2@il.qumranet.com> Sender: kvm-owner@vger.kernel.org List-ID: Le mardi 16 septembre 2008 =C3=A0 00:58 +0000, Avi Kivity a =C3=A9crit = : > From: Ben-Ami Yassour >=20 > Based on a patch by: Kay, Allen M >=20 > This patch enables PCI device assignment based on VT-d support. > When a device is assigned to the guest, the guest memory is pinned an= d > the mapping is updated in the VT-d IOMMU. >=20 > [Amit: Expose KVM_CAP_IOMMU so we can check if an IOMMU is present > and also control enable/disable from userspace] >=20 > Signed-off-by: Kay, Allen M > Signed-off-by: Weidong Han > Signed-off-by: Ben-Ami Yassour > Signed-off-by: Amit Shah >=20 > Acked-by: Mark Gross > Signed-off-by: Avi Kivity >=20 > diff --git a/arch/x86/kvm/Makefile b/arch/x86/kvm/Makefile > index d0e940b..3072b17 100644 > --- a/arch/x86/kvm/Makefile > +++ b/arch/x86/kvm/Makefile > @@ -12,6 +12,9 @@ EXTRA_CFLAGS +=3D -Ivirt/kvm -Iarch/x86/kvm > =20 > kvm-objs :=3D $(common-objs) x86.o mmu.o x86_emulate.o i8259.o irq.o= lapic.o \ > i8254.o > +ifeq ($(CONFIG_DMAR),y) > +kvm-objs +=3D vtd.o > +endif > obj-$(CONFIG_KVM) +=3D kvm.o > kvm-intel-objs =3D vmx.o > obj-$(CONFIG_KVM_INTEL) +=3D kvm-intel.o > diff --git a/arch/x86/kvm/vtd.c b/arch/x86/kvm/vtd.c > new file mode 100644 > index 0000000..667bf3f > --- /dev/null > +++ b/arch/x86/kvm/vtd.c > @@ -0,0 +1,198 @@ > +/* > + * Copyright (c) 2006, Intel Corporation. > + * > + * This program is free software; you can redistribute it and/or mod= ify it > + * under the terms and conditions of the GNU General Public License, > + * version 2, as published by the Free Software Foundation. > + * > + * This program is distributed in the hope it will be useful, but WI= THOUT > + * ANY WARRANTY; without even the implied warranty of MERCHANTABILIT= Y or > + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public Lic= ense for > + * more details. > + * > + * You should have received a copy of the GNU General Public License= along with > + * this program; if not, write to the Free Software Foundation, Inc.= , 59 Temple > + * Place - Suite 330, Boston, MA 02111-1307 USA. > + * > + * Copyright (C) 2006-2008 Intel Corporation > + * Copyright IBM Corporation, 2008 > + * Author: Allen M. Kay > + * Author: Weidong Han > + * Author: Ben-Ami Yassour > + */ > + > +#include > +#include > +#include > +#include > +#include > + > +static int kvm_iommu_unmap_memslots(struct kvm *kvm); > +static void kvm_iommu_put_pages(struct kvm *kvm, > + gfn_t base_gfn, unsigned long npages); > + > +int kvm_iommu_map_pages(struct kvm *kvm, > + gfn_t base_gfn, unsigned long npages) > +{ > + gfn_t gfn =3D base_gfn; > + pfn_t pfn; > + int i, r; > + struct dmar_domain *domain =3D kvm->arch.intel_iommu_domain; > + > + /* check if iommu exists and in use */ > + if (!domain) > + return 0; > + > + r =3D -EINVAL; > + for (i =3D 0; i < npages; i++) { > + /* check if already mapped */ > + pfn =3D (pfn_t)intel_iommu_iova_to_pfn(domain, > + gfn_to_gpa(gfn)); > + if (pfn && !is_mmio_pfn(pfn)) > + continue; > + > + pfn =3D gfn_to_pfn(kvm, gfn); > + if (!is_mmio_pfn(pfn)) { > + r =3D intel_iommu_page_mapping(domain, > + gfn_to_gpa(gfn), > + pfn_to_hpa(pfn), > + PAGE_SIZE, > + DMA_PTE_READ | > + DMA_PTE_WRITE); > + if (r) { > + printk(KERN_DEBUG "kvm_iommu_map_pages:" > + "iommu failed to map pfn=3D%lx\n", pfn); > + goto unmap_pages; > + } > + } else { > + printk(KERN_DEBUG "kvm_iommu_map_page:" > + "invalid pfn=3D%lx\n", pfn); > + goto unmap_pages; > + } > + gfn++; > + } > + return 0; > + > +unmap_pages: > + kvm_iommu_put_pages(kvm, base_gfn, i); > + return r; > +} > + > +static int kvm_iommu_map_memslots(struct kvm *kvm) > +{ > + int i, r; > + > + down_read(&kvm->slots_lock); > + for (i =3D 0; i < kvm->nmemslots; i++) { > + r =3D kvm_iommu_map_pages(kvm, kvm->memslots[i].base_gfn, > + kvm->memslots[i].npages); > + if (r) > + break; > + } > + up_read(&kvm->slots_lock); > + return r; > +} > + > +int kvm_iommu_map_guest(struct kvm *kvm, > + struct kvm_assigned_dev_kernel *assigned_dev) > +{ > + struct pci_dev *pdev =3D NULL; > + int r; > + > + if (!intel_iommu_found()) { > + printk(KERN_ERR "%s: intel iommu not found\n", __func__); > + return -ENODEV; > + } > + > + printk(KERN_DEBUG "VT-d direct map: host bdf =3D %x:%x:%x\n", > + assigned_dev->host_busnr, > + PCI_SLOT(assigned_dev->host_devfn), > + PCI_FUNC(assigned_dev->host_devfn)); > + > + pdev =3D assigned_dev->dev; > + > + if (pdev =3D=3D NULL) { > + if (kvm->arch.intel_iommu_domain) { > + intel_iommu_domain_exit(kvm->arch.intel_iommu_domain); > + kvm->arch.intel_iommu_domain =3D NULL; > + } > + return -ENODEV; > + } > + > + kvm->arch.intel_iommu_domain =3D intel_iommu_domain_alloc(pdev); > + if (!kvm->arch.intel_iommu_domain) > + return -ENODEV; > + > + r =3D kvm_iommu_map_memslots(kvm); > + if (r) > + goto out_unmap; > + > + intel_iommu_detach_dev(kvm->arch.intel_iommu_domain, > + pdev->bus->number, pdev->devfn); > + > + r =3D intel_iommu_context_mapping(kvm->arch.intel_iommu_domain, > + pdev); > + if (r) { > + printk(KERN_ERR "Domain context map for %s failed", > + pci_name(pdev)); > + goto out_unmap; > + } > + return 0; > + > +out_unmap: > + kvm_iommu_unmap_memslots(kvm); > + return r; > +} > + > +static void kvm_iommu_put_pages(struct kvm *kvm, > + gfn_t base_gfn, unsigned long npages) > +{ > + gfn_t gfn =3D base_gfn; > + pfn_t pfn; > + struct dmar_domain *domain =3D kvm->arch.intel_iommu_domain; > + int i; > + > + for (i =3D 0; i < npages; i++) { > + pfn =3D (pfn_t)intel_iommu_iova_to_pfn(domain, > + gfn_to_gpa(gfn)); > + kvm_release_pfn_clean(pfn); > + gfn++; > + } > +} > + > +static int kvm_iommu_unmap_memslots(struct kvm *kvm) > +{ > + int i; > + down_read(&kvm->slots_lock); > + for (i =3D 0; i < kvm->nmemslots; i++) { > + kvm_iommu_put_pages(kvm, kvm->memslots[i].base_gfn, > + kvm->memslots[i].npages); > + } > + up_read(&kvm->slots_lock); > + > + return 0; > +} > + > +int kvm_iommu_unmap_guest(struct kvm *kvm) > +{ > + struct kvm_assigned_dev_kernel *entry; > + struct dmar_domain *domain =3D kvm->arch.intel_iommu_domain; > + > + /* check if iommu exists and in use */ > + if (!domain) > + return 0; > + > + list_for_each_entry(entry, &kvm->arch.assigned_dev_head, list) { > + printk(KERN_DEBUG "VT-d unmap: host bdf =3D %x:%x:%x\n", > + entry->host_busnr, > + PCI_SLOT(entry->host_devfn), > + PCI_FUNC(entry->host_devfn)); > + > + /* detach kvm dmar domain */ > + intel_iommu_detach_dev(domain, entry->host_busnr, > + entry->host_devfn); > + } > + kvm_iommu_unmap_memslots(kvm); > + intel_iommu_domain_exit(domain); > + return 0; > +} > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > index 2134f3e..c8a2793 100644 > --- a/arch/x86/kvm/x86.c > +++ b/arch/x86/kvm/x86.c > @@ -35,6 +35,7 @@ > #include > #include > #include > +#include > =20 > #include > #include > @@ -277,9 +278,18 @@ static int kvm_vm_ioctl_assign_device(struct kvm= *kvm, > =20 > list_add(&match->list, &kvm->arch.assigned_dev_head); > =20 > + if (assigned_dev->flags & KVM_DEV_ASSIGN_ENABLE_IOMMU) { > + r =3D kvm_iommu_map_guest(kvm, match); > + if (r) > + goto out_list_del; > + } > + > out: > mutex_unlock(&kvm->lock); > return r; > +out_list_del: > + list_del(&match->list); > + pci_release_regions(dev); > out_disable: > pci_disable_device(dev); > out_put: > @@ -1147,6 +1157,9 @@ int kvm_dev_ioctl_check_extension(long ext) > case KVM_CAP_PV_MMU: > r =3D !tdp_enabled; > break; > + case KVM_CAP_IOMMU: > + r =3D intel_iommu_found(); > + break; Must depend on CONFIG_DMAR too Laurent --=20 ----------------- Laurent.Vivier@bull.net ------------------ "La perfection est atteinte non quand il ne reste rien =C3=A0 ajouter mais quand il ne reste rien =C3=A0 enlever." Saint Exup=C3=A9ry