From patchwork Thu Mar 1 10:33:35 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Liu, Yi L" X-Patchwork-Id: 879709 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: ozlabs.org; spf=pass (mailfrom) smtp.mailfrom=nongnu.org (client-ip=2001:4830:134:3::11; helo=lists.gnu.org; envelope-from=qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org; receiver=) Authentication-Results: ozlabs.org; dmarc=none (p=none dis=none) header.from=linux.intel.com Received: from lists.gnu.org (lists.gnu.org [IPv6:2001:4830:134:3::11]) (using TLSv1 with cipher AES256-SHA (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 3zsTs01FTvz9s06 for ; Thu, 1 Mar 2018 21:56:52 +1100 (AEDT) Received: from localhost ([::1]:55543 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1erLtJ-00026A-Cy for incoming@patchwork.ozlabs.org; Thu, 01 Mar 2018 05:56:49 -0500 Received: from eggs.gnu.org ([2001:4830:134:3::10]:53276) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1erLnZ-0005qF-V3 for qemu-devel@nongnu.org; Thu, 01 Mar 2018 05:50:55 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1erLnY-0000G6-JP for qemu-devel@nongnu.org; Thu, 01 Mar 2018 05:50:54 -0500 Received: from mga14.intel.com ([192.55.52.115]:7359) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1erLnY-00006Y-BG for qemu-devel@nongnu.org; Thu, 01 Mar 2018 05:50:52 -0500 X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga006.fm.intel.com ([10.253.24.20]) by fmsmga103.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 01 Mar 2018 02:50:51 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.47,408,1515484800"; d="scan'208";a="208048754" Received: from sky-dev.bj.intel.com ([10.238.144.105]) by fmsmga006.fm.intel.com with ESMTP; 01 Mar 2018 02:50:49 -0800 From: "Liu, Yi L" To: qemu-devel@nongnu.org, mst@redhat.com, david@gibson.dropbear.id.au Date: Thu, 1 Mar 2018 18:33:35 +0800 Message-Id: <1519900415-30314-13-git-send-email-yi.l.liu@linux.intel.com> X-Mailer: git-send-email 1.9.1 In-Reply-To: <1519900415-30314-1-git-send-email-yi.l.liu@linux.intel.com> References: <1519900415-30314-1-git-send-email-yi.l.liu@linux.intel.com> X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 192.55.52.115 Subject: [Qemu-devel] [PATCH v3 12/12] intel_iommu: bind device to PASID tagged AddressSpace X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: "Liu, Yi L" , kevin.tian@intel.com, yi.l.liu@intel.com, jasowang@redhat.com, peterx@redhat.com, alex.williamson@redhat.com, pbonzini@redhat.com, eric.auger.pro@gmail.com Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Sender: "Qemu-devel" This patch shows the idea of how a device is binded to a PASID tagged AddressSpace. when Intel vIOMMU emulator detected a pasid table entry programming from guest. Intel vIOMMU emulator firstly finds a VTDPASIDAddressSpace with the pasid field of pasid cache invalidate request. * If it is to bind a device to a guest process, needs add the device to the device list behind the VTDPASIDAddressSpace. And if the device is assigned device, need to register sva_notfier for future tlb flushing if any mapping changed to the process address space. * If it is to unbind a device from a guest process, then need to remove the device from the device list behind the VTDPASIDAddressSpace. And also needs to unregister the sva_notfier if the device is assigned device. This patch hasn't added the unbind logic. It depends on guest pasid table entry parsing which requires further emulation. Here just want to show the idea for the PASID tagged AddressSpace management framework. Full unregister logic would be included in future virt-SVA patchset. Signed-off-by: Liu, Yi L --- hw/i386/intel_iommu.c | 119 +++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 10 ++++ 2 files changed, 129 insertions(+) diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c index b8e8dbb..ed07035 100644 --- a/hw/i386/intel_iommu.c +++ b/hw/i386/intel_iommu.c @@ -1801,6 +1801,118 @@ static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) return true; } +static VTDPASIDAddressSpace *vtd_get_pasid_as(IntelIOMMUState *s, + uint32_t pasid) +{ + VTDPASIDAddressSpace *vtd_pasid_as = NULL; + IntelPASIDNode *node; + char name[128]; + + QLIST_FOREACH(node, &(s->pasid_as_list), next) { + vtd_pasid_as = node->pasid_as; + if (pasid == vtd_pasid_as->sva_ctx.pasid) { + return vtd_pasid_as; + } + } + + vtd_pasid_as = g_malloc0(sizeof(*vtd_pasid_as)); + vtd_pasid_as->iommu_state = s; + snprintf(name, sizeof(name), "intel_iommu_pasid_%d", pasid); + address_space_init(&vtd_pasid_as->as, NULL, "pasid"); + QLIST_INIT(&vtd_pasid_as->device_list); + + node = g_malloc0(sizeof(*node)); + node->pasid_as = vtd_pasid_as; + QLIST_INSERT_HEAD(&s->pasid_as_list, node, next); + + return vtd_pasid_as; +} + +static void vtd_bind_device_to_pasid_as(VTDPASIDAddressSpace *vtd_pasid_as, + PCIBus *bus, uint8_t devfn) +{ + VTDDeviceNode *node = NULL; + + QLIST_FOREACH(node, &(vtd_pasid_as->device_list), next) { + if (node->bus == bus && node->devfn == devfn) { + return; + } + } + + node = g_malloc0(sizeof(*node)); + node->bus = bus; + node->devfn = devfn; + QLIST_INSERT_HEAD(&(vtd_pasid_as->device_list), node, next); + + pci_device_sva_register_notifier(bus, devfn, &vtd_pasid_as->sva_ctx); + + return; +} + +static bool vtd_process_pc_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) +{ + + IntelIOMMUAssignedDeviceNode *node = NULL; + int ret = 0; + + uint16_t domain_id; + uint32_t pasid; + VTDPASIDAddressSpace *vtd_pasid_as; + + if ((inv_desc->lo & VTD_INV_DESC_PASIDC_RSVD_LO) || + (inv_desc->hi & VTD_INV_DESC_PASIDC_RSVD_HI)) { + return false; + } + + domain_id = VTD_INV_DESC_PASIDC_DID(inv_desc->lo); + + switch (inv_desc->lo & VTD_INV_DESC_PASIDC_G) { + case VTD_INV_DESC_PASIDC_ALL_ALL: + /* TODO: invalidate all pasid related cache */ + break; + + case VTD_INV_DESC_PASIDC_PASID_SI: + pasid = VTD_INV_DESC_PASIDC_PASID(inv_desc->lo); + vtd_pasid_as = vtd_get_pasid_as(s, pasid); + QLIST_FOREACH(node, &(s->assigned_device_list), next) { + VTDAddressSpace *vtd_as = node->vtd_as; + VTDContextEntry ce; + uint16_t did; + uint8_t bus = pci_bus_num(vtd_as->bus); + ret = vtd_dev_to_context_entry(s, bus, + vtd_as->devfn, &ce); + if (ret != 0) { + continue; + } + + did = VTD_CONTEXT_ENTRY_DID(ce.hi); + /* + * If did field equals to the domain_id field of inv_descriptor, + * then the device is affect by this invalidate request, need to + * bind or unbind the device to the pasid tagged address space. + * a) If it is bind, need to add the device to the device list, + * add register tlb flush notifier for it + * b) If it is unbind, need to remove the device from the device + * list, and unregister the tlb flush notifier + * TODO: add unbind logic accordingly, depends on the parsing of + * guest pasid table entry pasrsing, here has no parsing to + * pasid table entry. + * + */ + if (did == domain_id) { + vtd_bind_device_to_pasid_as(vtd_pasid_as, + vtd_as->bus, vtd_as->devfn); + } + } + break; + + default: + return false; + } + + return true; +} + static bool vtd_process_inv_iec_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) { @@ -1911,6 +2023,13 @@ static bool vtd_process_inv_desc(IntelIOMMUState *s) } break; + case VTD_INV_DESC_PC: + trace_vtd_inv_desc("pc", inv_desc.hi, inv_desc.lo); + if (!vtd_process_pc_desc(s, &inv_desc)) { + return false; + } + break; + case VTD_INV_DESC_IEC: trace_vtd_inv_desc("iec", inv_desc.hi, inv_desc.lo); if (!vtd_process_inv_iec_desc(s, &inv_desc)) { diff --git a/hw/i386/intel_iommu_internal.h b/hw/i386/intel_iommu_internal.h index d084099..31d0d53 100644 --- a/hw/i386/intel_iommu_internal.h +++ b/hw/i386/intel_iommu_internal.h @@ -332,6 +332,7 @@ typedef union VTDInvDesc VTDInvDesc; #define VTD_INV_DESC_IEC 0x4 /* Interrupt Entry Cache Invalidate Descriptor */ #define VTD_INV_DESC_WAIT 0x5 /* Invalidation Wait Descriptor */ +#define VTD_INV_DESC_PC 0x7 /* PASID-cache Invalidate Desc */ #define VTD_INV_DESC_NONE 0 /* Not an Invalidate Descriptor */ /* Masks for Invalidation Wait Descriptor*/ @@ -388,6 +389,15 @@ typedef union VTDInvDesc VTDInvDesc; #define VTD_SPTE_LPAGE_L4_RSVD_MASK(aw) \ (0x880ULL | ~(VTD_HAW_MASK(aw) | VTD_SL_IGN_COM)) +#define VTD_INV_DESC_PASIDC_G (3ULL << 4) +#define VTD_INV_DESC_PASIDC_PASID(val) (((val) >> 32) & 0xfffffULL) +#define VTD_INV_DESC_PASIDC_DID(val) (((val) >> 16) & VTD_DOMAIN_ID_MASK) +#define VTD_INV_DESC_PASIDC_RSVD_LO 0xfff000000000ffc0ULL +#define VTD_INV_DESC_PASIDC_RSVD_HI 0xffffffffffffffffULL + +#define VTD_INV_DESC_PASIDC_ALL_ALL (0ULL << 4) +#define VTD_INV_DESC_PASIDC_PASID_SI (1ULL << 4) + /* Information about page-selective IOTLB invalidate */ struct VTDIOTLBPageInvInfo { uint16_t domain_id;