intel_iommu: expose Shared Virtual Addressing to VMs
Subsystems (2)
v10 cover intel_iommu: expose Shared Virtual Addressing to VMs
Shared Virtual Addressing (SVA), a.k.a, Shared Virtual Memory (SVM) on Intel platforms allows address space sharing between device DMA and applications. SVA can reduce programming complexity and enhance security.
This QEMU series is intended to expose SVA usage to VMs. i.e. Sharing guest application address space with passthru devices. This is called vSVA in this series. The whole vSVA enabling requires QEMU/VFIO/IOMMU changes.
The high-level architecture for SVA virtualization is as below, the key design of vSVA support is to utilize the dual-stage IOMMU translation ( also known as IOMMU nesting translation) capability in host IOMMU.
.-------------. .---------------------------.
| vIOMMU | | Guest process CR3, FL only|
| | '---------------------------' .----------------/
| PASID Entry |--- PASID cache flush -
'-------------' |
| | V
| | CR3 in GPA
'-------------'
Guest
------| Shadow |--------------------------|--------
v v v
Host
.-------------. .----------------------.
| pIOMMU | | Bind FL for GVA-GPA |
| | '----------------------'
.----------------/ |
| PASID Entry | V (Nested xlate)
'----------------\.------------------------------.
| | |SL for GPA-HPA, default domain|
| | '------------------------------'
'-------------'Where:
- FL = First level/stage one page tables
- SL = Second level/stage two page tables
The complete vSVA kernel upstream patches are divided into three phases:
1. Common APIs and PCI device direct assignment
2. IOMMU-backed Mediated Device assignment
3. Page Request Services (PRS) support
This QEMU patchset is aiming for the phase 1 and phase 2. It is based on the kernel series below:
[PATCH v7 00/16] vfio: expose virtual Shared Virtual Addressing to VMs
https://lore.kernel.org/kvm/[email protected]/
Patch Overview:
1. patch 0001 - 0002: update kernel header files
2. patch 0003 - 0007: select VFIO_TYPE1_NESTING_IOMMU for vIOMMU built on IOMMU nesting translation.
3. patch 0008 - 0010: set HostIOMMUContext to vIOMMU.
4. patch 0011 - 0013: allocate PASID for vIOMMU.
5. patch 0014 - 0015: PASID cache management for Intel vIOMMU.
6. patch 0016 - 0020: bind guest page table to host.
7. patch 0021 - 0024: flush first level/stage cache for vIOMMU.
8. patch 0025: expose SVA to VM by x-scalable-mode="modern"
The complete QEMU set can be found in below link:
https://github.com/luxis1999/qemu.git vsva_5.9_rc2_qemu_rfcv10
Complete kernel can be found in:
https://github.com/luxis1999/linux-vsva.git vsva-linux-5.9-rc2-v7
Tests: basci vSVA functionality test, VM reboot/shutdown/crash, kernel build in guest, boot VM with vSVA disabled, full comapilation with all archs, passthru entire PCI device, passthru Scalable IOV ADI.
Regards,
Yi Liu
Changelog:
- RFC v9 -> RFC v10:
a) Rebase to latest kernel implementation (5.9-rc2 vsva v7)
RFC v9: https://lore.kernel.org/kvm/[email protected]/
- RFC v8 -> RFC v9:
a) Rebase to latest kernel implementation (5.8-rc6 vsva v6)
RFC v8: https://lore.kernel.org/kvm/[email protected]/
- RFC v7 -> RFC v8:
a) Rebase to latest kernel implementation (5.8-rc3 vsva v5)
RFC v7: https://lore.kernel.org/kvm/[email protected]/
- RFC v6 -> RFC v7:
a) Rebase to latest kernel implementation (5.8-rc3 vsva)
RFC v6: https://lore.kernel.org/kvm/[email protected]/
- RFC v5 -> RFC v6:
a) Use RFC instead of formal patch as kernel patch is in progress.
b) Address comments from Peter and Eric.
c) Add get_iommu_attr() to advertise vIOMMU nesting requirement to VFIO.
d) Update per latest kernel UAPI definition.
e) Add patch 0017 to check iommu nesting cap info in set_iommu(). RFC v5: https://www.spinics.net/lists/kvm/msg211475.html
- RFC v4 -> RFC v5:
a) Refactor the vfio HostIOMMUContext init code (patch 0008 - 0009 of v1 series)
b) Refactor the pasid binding handling (patch 0011 - 0016 of v1 series) RFC v4: https://patchwork.ozlabs.org/cover/1259648/
- RFC v3.1 -> RFC v4:
a) Implement HostIOMMUContext in QOM manner.
b) Add pci_set/unset_iommu_context() to register HostIOMMUContext to vIOMMU, thus the lifecircle of HostIOMMUContext is awared in vIOMMU side. In such way, vIOMMU could use the methods provided by the HostIOMMUContext safely.
c) Add back patch "[RFC v3 01/25] hw/pci: modify pci_setup_iommu() to set PCIIOMMUOps" RFCv3.1: https://patchwork.kernel.org/cover/11397879/
- RFC v3 -> v3.1:
a) Drop IOMMUContext, and rename DualStageIOMMUObject to HostIOMMUContext. HostIOMMUContext is per-vfio-container, it is exposed to vIOMMU via PCI layer. VFIO registers a PCIHostIOMMUFunc callback to PCI layer, vIOMMU could get HostIOMMUContext instance via it.
b) Check IOMMU uAPI version by VFIO_CHECK_EXTENSION
c) Add a check on VFIO_PASID_REQ availability via VFIO_GET_IOMMU_IHNFO
d) Reorder the series, put vSVA linux header file update in the beginning put the x-scalable-mode option mofification in the end of the series.
e) Dropped patch "[RFC v3 01/25] hw/pci: modify pci_setup_iommu() to set PCIIOMMUOps"
RFCv3: https://patchwork.kernel.org/cover/11356033/
- RFC v2 -> v3:
a) Introduce DualStageIOMMUObject to abstract the host IOMMU programming capability. e.g. request PASID from host, setup IOMMU nesting translation on host IOMMU. The pasid_alloc/bind_guest_page_table/iommu_cache_flush operations are moved to be DualStageIOMMUOps. Thus, DualStageIOMMUObject is an abstract layer which provides QEMU vIOMMU emulators with an explicit method to program host IOMMU.
b) Compared with RFC v2, the IOMMUContext has also been updated. It is modified to provide an abstract for vIOMMU emulators. It provides the method for pass-through modules (like VFIO) to communicate with host IOMMU. e.g. tell vIOMMU emulators about the IOMMU nesting capability on host side and report the host IOMMU DMA translation faults to vIOMMU emulators. RFC v2: https://www.spinics.net/lists/kvm/msg198556.html
- RFC v1 -> v2:
Introduce IOMMUContext to abstract the connection between VFIO and vIOMMU emulators, which is a replacement of the PCIPASIDOps in RFC v1. Modify x-scalable-mode to be string option instead of adding a new option as RFC v1 did. Refined the pasid cache management
---
Eric Auger (1):
scripts/update-linux-headers: Import iommu.h
Liu Yi L (24):
header file update VFIO/IOMMU vSVA APIs kernel 5.9-rc2 hw/pci: modify pci_setup_iommu() to set PCIIOMMUOps hw/pci: introduce pci_device_get_iommu_attr()
intel_iommu: add get_iommu_attr() callback
vfio: pass nesting requirement into vfio_get_group()
vfio: check VFIO_TYPE1_NESTING_IOMMU support hw/iommu: introduce HostIOMMUContext
hw/pci: introduce pci_device_set/unset_iommu_context()
intel_iommu: add set/unset_iommu_context callback vfio/common: provide PASID alloc/free hooks
vfio: init HostIOMMUContext per-container
intel_iommu: add virtual command capability support
intel_iommu: process PASID cache invalidation
intel_iommu: add PASID cache management infrastructure
vfio: add bind stage-1 page table support
intel_iommu: sync IOMMU nesting cap info for assigned devices
intel_iommu: bind/unbind guest page table to host
intel_iommu: replay pasid binds after context cache invalidation
intel_iommu: do not pass down pasid bind for PASID #0
vfio: add support for flush iommu stage-1 cache
intel_iommu: process PASID-based iotlb invalidation
intel_iommu: propagate PASID-based iotlb invalidation to host
intel_iommu: process PASID-based Device-TLB invalidation
intel_iommu: modify x-scalable-mode to be string option
hw/Kconfig | 3 + hw/alpha/typhoon.c | 6 +- hw/arm/smmu-common.c | 6 +- hw/hppa/dino.c | 6 +- hw/i386/amd_iommu.c | 6 +- hw/i386/intel_iommu.c | 1231 ++++++++++++++++++++++++++++++++- hw/i386/intel_iommu_internal.h | 131 ++++ hw/i386/trace-events | 6 + hw/iommu/Kconfig | 4 + hw/iommu/host_iommu_context.c | 171 +++++ hw/iommu/meson.build | 6 + hw/meson.build | 1 + hw/pci-host/designware.c | 6 +- hw/pci-host/pnv_phb3.c | 6 +- hw/pci-host/pnv_phb4.c | 6 +- hw/pci-host/ppce500.c | 6 +- hw/pci-host/prep.c | 6 +- hw/pci-host/sabre.c | 6 +- hw/pci/pci.c | 73 +- hw/ppc/ppc440_pcix.c | 6 +- hw/ppc/spapr_pci.c | 6 +- hw/s390x/s390-pci-bus.c | 8 +- hw/vfio/ap.c | 2 +- hw/vfio/ccw.c | 2 +- hw/vfio/common.c | 299 +++++++- hw/vfio/pci.c | 26 +- hw/vfio/platform.c | 2 +- hw/virtio/virtio-iommu.c | 6 +- include/hw/i386/intel_iommu.h | 61 +- include/hw/iommu/host_iommu_context.h | 103 +++ include/hw/pci/pci.h | 25 +- include/hw/pci/pci_bus.h | 2 +- include/hw/vfio/vfio-common.h | 7 +- linux-headers/linux/iommu.h | 420 +++++++++++ linux-headers/linux/vfio.h | 103 ++- scripts/update-linux-headers.sh | 2 +- 36 files changed, 2705 insertions(+), 61 deletions(-) create mode 100644 hw/iommu/Kconfig create mode 100644 hw/iommu/host_iommu_context.c create mode 100644 hw/iommu/meson.build create mode 100644 include/hw/iommu/host_iommu_context.h create mode 100644 linux-headers/linux/iommu.h
Signature · 1 line
2.7.4
RFC v10 1/25 scripts/update-linux-headers: Import iommu.h
From: Eric Auger <[email protected]>
Update the script to import the new iommu.h uapi header.
--- scripts/update-linux-headers.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
scripts/update-linux-headers.sh +1 -1
@@ -141,7 +141,7 @@ done rm -rf "$output/linux-headers/linux" mkdir -p "$output/linux-headers/linux" -for header in kvm.h vfio.h vfio_ccw.h vhost.h \ +for header in kvm.h vfio.h vfio_ccw.h vhost.h iommu.h \ psci.h psp-sev.h userfaultfd.h mman.h; do cp "$tmpdir/include/linux/$header" "$output/linux-headers/linux" done
Signature · 1 line
2.7.4
RFC v10 2/25 header file update VFIO/IOMMU vSVA APIs kernel 5.9-rc2
In reply to Liu Yi LThe kernel uapi/linux/iommu.h header file includes the extensions for vSVA support. e.g. bind gpasid, iommu fault report related user structures and etc.
This commit updates kernel headers from the below branch:
https://github.com/luxis1999/linux-vsva.git vsva-linux-5.9-rc2-v7
Note: this should be replaced with a full header files update when the vSVA uPAPI is stable.
--- linux-headers/linux/iommu.h | 420 ++++++++++++++++++++++++++++++++++++++++++++ linux-headers/linux/vfio.h | 103 ++++++++++- 2 files changed, 522 insertions(+), 1 deletion(-) create mode 100644 linux-headers/linux/iommu.h
linux-headers/linux/iommu.h new file +420 -0
@@ -0,0 +1,420 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +/* +* IOMMU user API definitions +*/ + +#ifndef _IOMMU_H +#define _IOMMU_H + +#include <linux/types.h> + +#define IOMMU_FAULT_PERM_READ (1 << 0)/* read */ +#define IOMMU_FAULT_PERM_WRITE (1 << 1)/* write */ +#define IOMMU_FAULT_PERM_EXEC (1 << 2)/* exec */ +#define IOMMU_FAULT_PERM_PRIV (1 << 3)/* privileged */ + +/* Generic fault types, can be expanded IRQ remapping fault */ +enum iommu_fault_type { +IOMMU_FAULT_DMA_UNRECOV =1 ,/* unrecoverable fault */ +IOMMU_FAULT_PAGE_REQ ,/* page request fault */ +}; + +enum iommu_fault_reason { +IOMMU_FAULT_REASON_UNKNOWN =0 , + +/* Could not access the PASID table (fetch caused external abort) */ +IOMMU_FAULT_REASON_PASID_FETCH , + +/* PASID entry is invalid or has configuration errors */ +IOMMU_FAULT_REASON_BAD_PASID_ENTRY , + +/* +* PASID is out of range (e.g. exceeds the maximum PASID +* supported by the IOMMU) or disabled. +*/ +IOMMU_FAULT_REASON_PASID_INVALID , + +/* +* An external abort occurred fetching (or updating) a translation +* table descriptor +*/ +IOMMU_FAULT_REASON_WALK_EABT , + +/* +* Could not access the page table entry (Bad address), +* actual translation fault +*/ +IOMMU_FAULT_REASON_PTE_FETCH , + +/* Protection flag check failed */ +IOMMU_FAULT_REASON_PERMISSION , + +/* access flag check failed */ +IOMMU_FAULT_REASON_ACCESS , + +/* Output address of a translation stage caused Address Size fault */ +IOMMU_FAULT_REASON_OOR_ADDRESS , +}; + +/** +* struct iommu_fault_unrecoverable - Unrecoverable fault data +* @reason: reason of the fault, from &enum iommu_fault_reason +* @flags: parameters of this fault (IOMMU_FAULT_UNRECOV_* values) +* @pasid: Process Address Space ID +* @perm: requested permission access using by the incoming transaction +* (IOMMU_FAULT_PERM_* values) +* @addr: offending page address +* @fetch_addr: address that caused a fetch abort, if any +*/ +struct iommu_fault_unrecoverable { +__u32 reason; +#define IOMMU_FAULT_UNRECOV_PASID_VALID (1 << 0) +#define IOMMU_FAULT_UNRECOV_ADDR_VALID (1 << 1) +#define IOMMU_FAULT_UNRECOV_FETCH_ADDR_VALID (1 << 2) +__u32 flags; +__u32 pasid; +__u32 perm; +__u64 addr; +__u64 fetch_addr; +}; + +/** +* struct iommu_fault_page_request - Page Request data +* @flags: encodes whether the corresponding fields are valid and whether this +* is the last page in group (IOMMU_FAULT_PAGE_REQUEST_* values). +* When IOMMU_FAULT_PAGE_RESPONSE_NEEDS_PASID is set, the page response +* must have the same PASID value as the page request. When it is clear, +* the page response should not have a PASID. +* @pasid: Process Address Space ID +* @grpid: Page Request Group Index +* @perm: requested page permissions (IOMMU_FAULT_PERM_* values) +* @addr: page address +* @private_data: device-specific private information +*/ +struct iommu_fault_page_request { +#define IOMMU_FAULT_PAGE_REQUEST_PASID_VALID (1 << 0) +#define IOMMU_FAULT_PAGE_REQUEST_LAST_PAGE (1 << 1) +#define IOMMU_FAULT_PAGE_REQUEST_PRIV_DATA (1 << 2) +#define IOMMU_FAULT_PAGE_RESPONSE_NEEDS_PASID (1 << 3) +__u32 flags; +__u32 pasid; +__u32 grpid; +__u32 perm; +__u64 addr; +__u64 private_data[2 ]; +}; + +/** +* struct iommu_fault - Generic fault data +* @type: fault type from &enum iommu_fault_type +* @padding: reserved for future use (should be zero) +* @event: fault event, when @type is %IOMMU_FAULT_DMA_UNRECOV +* @prm: Page Request message, when @type is %IOMMU_FAULT_PAGE_REQ +* @padding2: sets the fault size to allow for future extensions +*/ +struct iommu_fault { +__u32 type; +__u32 padding; +union { +struct iommu_fault_unrecoverable event; +struct iommu_fault_page_request prm; +__u8 padding2[56 ]; + }; +}; + +/** +* enum iommu_page_response_code - Return status of fault handlers +* @IOMMU_PAGE_RESP_SUCCESS: Fault has been handled and the page tables +* populated, retry the access. This is "Success" in PCI PRI. +* @IOMMU_PAGE_RESP_FAILURE: General error. Drop all subsequent faults from +* this device if possible. This is "Response Failure" in PCI PRI. +* @IOMMU_PAGE_RESP_INVALID: Could not handle this fault, don't retry the +* access. This is "Invalid Request" in PCI PRI. +*/ +enum iommu_page_response_code { +IOMMU_PAGE_RESP_SUCCESS =0 , +IOMMU_PAGE_RESP_INVALID , +IOMMU_PAGE_RESP_FAILURE , +}; + +/** +* struct iommu_page_response - Generic page response information +* @argsz: User filled size of this data +* @version: API version of this structure +* @flags: encodes whether the corresponding fields are valid +* (IOMMU_FAULT_PAGE_RESPONSE_* values) +* @pasid: Process Address Space ID +* @grpid: Page Request Group Index +* @code: response code from &enum iommu_page_response_code +*/ +struct iommu_page_response { +__u32 argsz; +#define IOMMU_PAGE_RESP_VERSION_1 1 +__u32 version; +#define IOMMU_PAGE_RESP_PASID_VALID (1 << 0) +__u32 flags; +__u32 pasid; +__u32 grpid; +__u32 code; +}; + +/* defines the granularity of the invalidation */ +enum iommu_inv_granularity { +IOMMU_INV_GRANU_DOMAIN ,/* domain-selective invalidation */ +IOMMU_INV_GRANU_PASID ,/* PASID-selective invalidation */ +IOMMU_INV_GRANU_ADDR ,/* page-selective invalidation */ +IOMMU_INV_GRANU_NR ,/* number of invalidation granularities */ +}; + +/** +* struct iommu_inv_addr_info - Address Selective Invalidation Structure +* +* @flags: indicates the granularity of the address-selective invalidation +* - If the PASID bit is set, the @pasid field is populated and the invalidation +* relates to cache entries tagged with this PASID and matching the address +* range. +* - If ARCHID bit is set, @archid is populated and the invalidation relates +* to cache entries tagged with this architecture specific ID and matching +* the address range. +* - Both PASID and ARCHID can be set as they may tag different caches. +* - If neither PASID or ARCHID is set, global addr invalidation applies. +* - The LEAF flag indicates whether only the leaf PTE caching needs to be +* invalidated and other paging structure caches can be preserved. +* @pasid: process address space ID +* @archid: architecture-specific ID +* @addr: first stage/level input address +* @granule_size: page/block size of the mapping in bytes +* @nb_granules: number of contiguous granules to be invalidated +*/ +struct iommu_inv_addr_info { +#define IOMMU_INV_ADDR_FLAGS_PASID (1 << 0) +#define IOMMU_INV_ADDR_FLAGS_ARCHID (1 << 1) +#define IOMMU_INV_ADDR_FLAGS_LEAF (1 << 2) +__u32 flags; +__u32 archid; +__u64 pasid; +__u64 addr; +__u64 granule_size; +__u64 nb_granules; +}; + +/** +* struct iommu_inv_pasid_info - PASID Selective Invalidation Structure +* +* @flags: indicates the granularity of the PASID-selective invalidation +* - If the PASID bit is set, the @pasid field is populated and the invalidation +* relates to cache entries tagged with this PASID and matching the address +* range. +* - If the ARCHID bit is set, the @archid is populated and the invalidation +* relates to cache entries tagged with this architecture specific ID and +* matching the address range. +* - Both PASID and ARCHID can be set as they may tag different caches. +* - At least one of PASID or ARCHID must be set. +* @pasid: process address space ID +* @archid: architecture-specific ID +*/ +struct iommu_inv_pasid_info { +#define IOMMU_INV_PASID_FLAGS_PASID (1 << 0) +#define IOMMU_INV_PASID_FLAGS_ARCHID (1 << 1) +__u32 flags; +__u32 archid; +__u64 pasid; +}; + +/** +* struct iommu_cache_invalidate_info - First level/stage invalidation +* information +* @argsz: User filled size of this data +* @version: API version of this structure +* @cache: bitfield that allows to select which caches to invalidate +* @granularity: defines the lowest granularity used for the invalidation: +* domain > PASID > addr +* @padding: reserved for future use (should be zero) +* @pasid_info: invalidation data when @granularity is %IOMMU_INV_GRANU_PASID +* @addr_info: invalidation data when @granularity is %IOMMU_INV_GRANU_ADDR +* +* Not all the combinations of cache/granularity are valid: +* +* +--------------+---------------+---------------+---------------+ +* | type / | DEV_IOTLB | IOTLB | PASID | +* | granularity | | | cache | +* +==============+===============+===============+===============+ +* | DOMAIN | N/A | Y | Y | +* +--------------+---------------+---------------+---------------+ +* | PASID | Y | Y | Y | +* +--------------+---------------+---------------+---------------+ +* | ADDR | Y | Y | N/A | +* +--------------+---------------+---------------+---------------+ +* +* Invalidations by %IOMMU_INV_GRANU_DOMAIN don't take any argument other than +* @version and @cache. +* +* If multiple cache types are invalidated simultaneously, they all +* must support the used granularity. +*/ +struct iommu_cache_invalidate_info { +__u32 argsz; +#define IOMMU_CACHE_INVALIDATE_INFO_VERSION_1 1 +__u32 version; +/* IOMMU paging structure cache */ +#define IOMMU_CACHE_INV_TYPE_IOTLB (1 << 0)/* IOMMU IOTLB */ +#define IOMMU_CACHE_INV_TYPE_DEV_IOTLB (1 << 1)/* Device IOTLB */ +#define IOMMU_CACHE_INV_TYPE_PASID (1 << 2)/* PASID cache */ +#define IOMMU_CACHE_INV_TYPE_NR (3) +__u8 cache; +__u8 granularity; +__u8 padding[6 ]; +union { +struct iommu_inv_pasid_info pasid_info; +struct iommu_inv_addr_info addr_info; + } granu; +}; + +/** +* struct iommu_gpasid_bind_data_vtd - Intel VT-d specific data on device and guest +* SVA binding. +* +* @flags: VT-d PASID table entry attributes +* @pat: Page attribute table data to compute effective memory type +* @emt: Extended memory type +* +* Only guest vIOMMU selectable and effective options are passed down to +* the host IOMMU. +*/ +struct iommu_gpasid_bind_data_vtd { +#define IOMMU_SVA_VTD_GPASID_SRE (1 << 0)/* supervisor request */ +#define IOMMU_SVA_VTD_GPASID_EAFE (1 << 1)/* extended access enable */ +#define IOMMU_SVA_VTD_GPASID_PCD (1 << 2)/* page-level cache disable */ +#define IOMMU_SVA_VTD_GPASID_PWT (1 << 3)/* page-level write through */ +#define IOMMU_SVA_VTD_GPASID_EMTE (1 << 4)/* extended mem type enable */ +#define IOMMU_SVA_VTD_GPASID_CD (1 << 5)/* PASID-level cache disable */ +#define IOMMU_SVA_VTD_GPASID_LAST (1 << 6) +__u64 flags; +__u32 pat; +__u32 emt; +}; + +#define IOMMU_SVA_VTD_GPASID_MTS_MASK (IOMMU_SVA_VTD_GPASID_CD | \ + IOMMU_SVA_VTD_GPASID_EMTE | \ + IOMMU_SVA_VTD_GPASID_PCD | \ + IOMMU_SVA_VTD_GPASID_PWT) + +enum iommu_pasid_data_format { +IOMMU_PASID_FORMAT_INTEL_VTD =1 , +IOMMU_PASID_FORMAT_LAST , +}; + +/** +* struct iommu_gpasid_bind_data - Information about device and guest PASID binding +* @argsz: User filled size of this data +* @version: Version of this data structure +* @format: PASID table entry format of enum iommu_pasid_data_format type +* @flags: Additional information on guest bind request +* @gpgd: Guest page directory base of the guest mm to bind +* @hpasid: Process address space ID used for the guest mm in host IOMMU +* @gpasid: Process address space ID used for the guest mm in guest IOMMU +* @addr_width: Guest virtual address width +* @padding: Reserved for future use (should be zero) +* @vtd: Intel VT-d specific data +* +* Guest to host PASID mapping can be an identity or non-identity, where guest +* has its own PASID space. For non-identify mapping, guest to host PASID lookup +* is needed when VM programs guest PASID into an assigned device. VMM may +* trap such PASID programming then request host IOMMU driver to convert guest +* PASID to host PASID based on this bind data. +*/ +struct iommu_gpasid_bind_data { +__u32 argsz; +#define IOMMU_GPASID_BIND_VERSION_1 1 +__u32 version; +__u32 format; +__u32 addr_width; +#define IOMMU_SVA_GPASID_VAL (1 << 0)/* guest PASID valid */ +__u64 flags; +__u64 gpgd; +__u64 hpasid; +__u64 gpasid; +__u8 padding[8 ]; +/* Vendor specific data */ +union { +struct iommu_gpasid_bind_data_vtd vtd; + } vendor; +}; + +/* +* struct iommu_nesting_info_vtd - Intel VT-d specific nesting info. +* +* @flags: VT-d specific flags. Currently reserved for future +* extension. must be set to 0. +* @cap_reg: Describe basic capabilities as defined in VT-d capability +* register. +* @ecap_reg: Describe the extended capabilities as defined in VT-d +* extended capability register. +*/ +struct iommu_nesting_info_vtd { +__u32 flags; +__u64 cap_reg; +__u64 ecap_reg; +}; + +/* +* struct iommu_nesting_info - Information for nesting-capable IOMMU. +* userspace should check it before using +* nesting capability. +* +* @argsz: size of the whole structure. +* @flags: currently reserved for future extension. must set to 0. +* @format: PASID table entry format, the same definition as struct +* iommu_gpasid_bind_data @format. +* @features: supported nesting features. +* @addr_width: The output addr width of first level/stage translation +* @pasid_bits: Maximum supported PASID bits, 0 represents no PASID +* support. +* @vendor: vendor specific data, structure type can be deduced from +* @format field. +* +* +===============+======================================================+ +* | feature | Notes | +* +===============+======================================================+ +* | SYSWIDE_PASID | IOMMU vendor driver sets it to mandate userspace | +* | | to allocate PASID from kernel. All PASID allocation | +* | | free must be mediated through the IOMMU UAPI. | +* +---------------+------------------------------------------------------+ +* | BIND_PGTBL | IOMMU vendor driver sets it to mandate userspace to | +* | | bind the first level/stage page table to associated | +* | | PASID (either the one specified in bind request or | +* | | the default PASID of iommu domain), through IOMMU | +* | | UAPI. | +* +---------------+------------------------------------------------------+ +* | CACHE_INVLD | IOMMU vendor driver sets it to mandate userspace to | +* | | explicitly invalidate the IOMMU cache through IOMMU | +* | | UAPI according to vendor-specific requirement when | +* | | changing the 1st level/stage page table. | +* +---------------+------------------------------------------------------+ +* +* data struct types defined for @format: +* +================================+=====================================+ +* | @format | data struct | +* +================================+=====================================+ +* | IOMMU_PASID_FORMAT_INTEL_VTD | struct iommu_nesting_info_vtd | +* +--------------------------------+-------------------------------------+ +* +*/ +struct iommu_nesting_info { +__u32 argsz; +__u32 flags; +__u32 format; +#define IOMMU_NESTING_FEAT_SYSWIDE_PASID (1 << 0) +#define IOMMU_NESTING_FEAT_BIND_PGTBL (1 << 1) +#define IOMMU_NESTING_FEAT_CACHE_INVLD (1 << 2) +__u32 features; +__u16 addr_width; +__u16 pasid_bits; +__u8 padding[12 ]; +/* Vendor specific data */ +union { +struct iommu_nesting_info_vtd vtd; + } vendor; +}; + +#endif /* _IOMMU_H */
linux-headers/linux/vfio.h +102 -1
127 lines not on this page · raw message
Signature · 1 line
2.7.4
RFC v10 3/25 hw/pci: modify pci_setup_iommu() to set PCIIOMMUOps
In reply to Liu Yi LThis patch modifies pci_setup_iommu() to set PCIIOMMUOps instead of setting PCIIOMMUFunc. PCIIOMMUFunc is used to get an address space for a PCI device in vendor specific way. The PCIIOMMUOps still offers this functionality. But using PCIIOMMUOps leaves space to add more iommu related vendor specific operations.
---
rfcv9 -> rfcv10:
*) Fix a bug in pci_device_iommu_address_space()
+ iommu_bus->iommu_ops->get_address_space) &&
=>
+ !iommu_bus->iommu_ops->get_address_space) &&
--- hw/alpha/typhoon.c | 6 +++++- hw/arm/smmu-common.c | 6 +++++- hw/hppa/dino.c | 6 +++++- hw/i386/amd_iommu.c | 6 +++++- hw/i386/intel_iommu.c | 6 +++++- hw/pci-host/designware.c | 6 +++++- hw/pci-host/pnv_phb3.c | 6 +++++- hw/pci-host/pnv_phb4.c | 6 +++++- hw/pci-host/ppce500.c | 6 +++++- hw/pci-host/prep.c | 6 +++++- hw/pci-host/sabre.c | 6 +++++- hw/pci/pci.c | 18 +++++++++++++----- hw/ppc/ppc440_pcix.c | 6 +++++- hw/ppc/spapr_pci.c | 6 +++++- hw/s390x/s390-pci-bus.c | 8 ++++++-- hw/virtio/virtio-iommu.c | 6 +++++- include/hw/pci/pci.h | 8 ++++++-- include/hw/pci/pci_bus.h | 2 +- 18 files changed, 96 insertions(+), 24 deletions(-)
hw/alpha/typhoon.c +5 -1
@@ -740,6 +740,10 @@ static AddressSpace *typhoon_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &s->pchip.iommu_as; } +static const PCIIOMMUOps typhoon_iommu_ops = { + .get_address_space = typhoon_pci_dma_iommu, +}; +static void typhoon_set_irq (void *opaque,int irq,int level) {TyphoonState *s = opaque;
@@ -897,7 +901,7 @@ PCIBus *typhoon_init(MemoryRegion *ram, ISABus **isa_bus, qemu_irq *p_rtc_irq, "iommu-typhoon", UINT64_MAX);address_space_init (&s->pchip.iommu_as,MEMORY_REGION (&s->pchip.iommu),"pchip0-pci" ); -pci_setup_iommu (b, typhoon_pci_dma_iommu, s); +pci_setup_iommu (b, &typhoon_iommu_ops, s);/* Pchip0 PCI special/interrupt acknowledge, 0x801.F800.0000, 64MB. */ memory_region_init_io(&s->pchip.reg_iack, OBJECT(s), &alpha_pci_iack_ops,
hw/arm/smmu-common.c +5 -1
@@ -444,6 +444,10 @@ static AddressSpace *smmu_find_add_as(PCIBus *bus, void *opaque, int devfn)return &sdev->as; } +static const PCIIOMMUOps smmu_ops = { + .get_address_space = smmu_find_add_as, +}; +IOMMUMemoryRegion *smmu_iommu_mr (SMMUState *s,uint32_t sid) {uint8_t bus_n, devfn;
@@ -513,7 +517,7 @@ static void smmu_base_realize(DeviceState *dev, Error **errp) s->smmu_pcibus_by_busptr =g_hash_table_new (NULL ,NULL );if (s->primary_bus) { -pci_setup_iommu (s->primary_bus, smmu_find_add_as, s); +pci_setup_iommu (s->primary_bus, &smmu_ops, s); }else {error_setg (errp,"SMMU is not attached to any PCI bus!" ); }
hw/hppa/dino.c +5 -1
@@ -459,6 +459,10 @@ static AddressSpace *dino_pcihost_set_iommu(PCIBus *bus, void *opaque,return &s->bm_as; } +static const PCIIOMMUOps dino_iommu_ops = { + .get_address_space = dino_pcihost_set_iommu, +}; +/* * Dino interrupts are connected as shown on Page 78, Table 23 * (Little-endian bit numbers)
@@ -580,7 +584,7 @@ PCIBus *dino_init(MemoryRegion *addr_space,memory_region_add_subregion (&s->bm,0xfff00000 , &s->bm_cpu_alias);address_space_init (&s->bm_as, &s->bm,"pci-bm" ); -pci_setup_iommu (b, dino_pcihost_set_iommu, s); +pci_setup_iommu (b, &dino_iommu_ops, s); *p_rtc_irq =qemu_allocate_irq (dino_set_timer_irq, s,0 ); *p_ser_irq =qemu_allocate_irq (dino_set_serial_irq, s,0 );
hw/i386/amd_iommu.c +5 -1
@@ -1452,6 +1452,10 @@ static AddressSpace *amdvi_host_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &iommu_as[devfn]->as; } +static const PCIIOMMUOps amdvi_iommu_ops = { + .get_address_space = amdvi_host_dma_iommu, +}; +static const MemoryRegionOps mmio_mem_ops = { .read = amdvi_mmio_read, .write = amdvi_mmio_write,
@@ -1579,7 +1583,7 @@ static void amdvi_realize(DeviceState *dev, Error **errp)sysbus_init_mmio (SYS_BUS_DEVICE (s), &s->mmio);sysbus_mmio_map (SYS_BUS_DEVICE (s),0 ,AMDVI_BASE_ADDR ); -pci_setup_iommu (bus, amdvi_host_dma_iommu, s); +pci_setup_iommu (bus, &amdvi_iommu_ops, s); s->devid =object_property_get_int (OBJECT (&s->pci),"addr" , &error_abort);msi_init (&s->pci.dev,0 ,1 , true, false, errp);amdvi_init (s);
hw/i386/intel_iommu.c +5 -1
@@ -3733,6 +3733,10 @@ static AddressSpace *vtd_host_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &vtd_as->as; } +static PCIIOMMUOps vtd_iommu_ops = { + .get_address_space = vtd_host_dma_iommu, +}; +static bool vtd_decide_config (IntelIOMMUState *s,Error **errp) {X86IOMMUState *x86_iommu =X86_IOMMU_DEVICE (s);
@@ -3844,7 +3848,7 @@ static void vtd_realize(DeviceState *dev, Error **errp) g_free, g_free);vtd_init (s);sysbus_mmio_map (SYS_BUS_DEVICE (s),0 ,Q35_HOST_BRIDGE_IOMMU_ADDR ); -pci_setup_iommu (bus, vtd_host_dma_iommu, dev); +pci_setup_iommu (bus, &vtd_iommu_ops, dev);/* Pseudo address space under root PCI bus. */ x86ms->ioapic_as =vtd_host_dma_iommu (bus, s,Q35_PSEUDO_DEVFN_IOAPIC );qemu_add_machine_init_done_notifier (&vtd_machine_done_notify);
hw/pci-host/designware.c +5 -1
@@ -645,6 +645,10 @@ static AddressSpace *designware_pcie_host_set_iommu(PCIBus *bus, void *opaque,return &s->pci.address_space; } +static const PCIIOMMUOps designware_iommu_ops = { + .get_address_space = designware_pcie_host_set_iommu, +}; +static void designware_pcie_host_realize (DeviceState *dev,Error **errp) {PCIHostState *pci =PCI_HOST_BRIDGE (dev);
@@ -686,7 +690,7 @@ static void designware_pcie_host_realize(DeviceState *dev, Error **errp)address_space_init (&s->pci.address_space, &s->pci.address_space_root,"pcie-bus-address-space" ); -pci_setup_iommu (pci->bus, designware_pcie_host_set_iommu, s); +pci_setup_iommu (pci->bus, &designware_iommu_ops, s);qdev_realize (DEVICE (&s->root),BUS (pci->bus), &error_fatal); }
hw/pci-host/pnv_phb3.c +5 -1
@@ -961,6 +961,10 @@ static AddressSpace *pnv_phb3_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &ds->dma_as; } +static PCIIOMMUOps pnv_phb3_iommu_ops = { + .get_address_space = pnv_phb3_dma_iommu, +}; +static void pnv_phb3_instance_init (Object *obj) {PnvPHB3 *phb =PNV_PHB3 (obj);
@@ -1048,7 +1052,7 @@ static void pnv_phb3_realize(DeviceState *dev, Error **errp) &phb->pci_mmio, &phb->pci_io, 0, 4, TYPE_PNV_PHB3_ROOT_BUS); -pci_setup_iommu (pci->bus, pnv_phb3_dma_iommu, phb); +pci_setup_iommu (pci->bus, &pnv_phb3_iommu_ops, phb);/* Add a single Root port */ qdev_prop_set_uint8 (DEVICE (&phb->root),"chassis" , phb->chip_id);
hw/pci-host/pnv_phb4.c +5 -1
@@ -1148,6 +1148,10 @@ static AddressSpace *pnv_phb4_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &ds->dma_as; } +static PCIIOMMUOps pnv_phb4_iommu_ops = { + .get_address_space = pnv_phb4_dma_iommu, +}; +static void pnv_phb4_instance_init (Object *obj) {PnvPHB4 *phb =PNV_PHB4 (obj);
@@ -1202,7 +1206,7 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp) pnv_phb4_set_irq, pnv_phb4_map_irq, phb, &phb->pci_mmio, &phb->pci_io, 0, 4, TYPE_PNV_PHB4_ROOT_BUS); -pci_setup_iommu (pci->bus, pnv_phb4_dma_iommu, phb); +pci_setup_iommu (pci->bus, &pnv_phb4_iommu_ops, phb);/* Add a single Root port */ qdev_prop_set_uint8 (DEVICE (&phb->root),"chassis" , phb->chip_id);
hw/pci-host/ppce500.c +5 -1
@@ -439,6 +439,10 @@ static AddressSpace *e500_pcihost_set_iommu(PCIBus *bus, void *opaque,return &s->bm_as; } +static const PCIIOMMUOps ppce500_iommu_ops = { + .get_address_space = e500_pcihost_set_iommu, +}; +static void e500_pcihost_realize (DeviceState *dev,Error **errp) {SysBusDevice *sbd =SYS_BUS_DEVICE (dev);
@@ -473,7 +477,7 @@ static void e500_pcihost_realize(DeviceState *dev, Error **errp)memory_region_init (&s->bm,OBJECT (s),"bm-e500" ,UINT64_MAX );memory_region_add_subregion (&s->bm,0x0 , &s->busmem);address_space_init (&s->bm_as, &s->bm,"pci-bm" ); -pci_setup_iommu (b, e500_pcihost_set_iommu, s); +pci_setup_iommu (b, &ppce500_iommu_ops, s);pci_create_simple (b,0 ,"e500-host-bridge" );
hw/pci-host/prep.c +5 -1
@@ -213,6 +213,10 @@ static AddressSpace *raven_pcihost_set_iommu(PCIBus *bus, void *opaque,return &s->bm_as; } +static const PCIIOMMUOps raven_iommu_ops = { + .get_address_space = raven_pcihost_set_iommu, +}; +static void raven_change_gpio (void *opaque,int n,int level) {PREPPCIState *s = opaque;
@@ -301,7 +305,7 @@ static void raven_pcihost_initfn(Object *obj)memory_region_add_subregion (&s->bm,0 , &s->bm_pci_memory_alias);memory_region_add_subregion (&s->bm,0x80000000 , &s->bm_ram_alias);address_space_init (&s->bm_as, &s->bm,"raven-bm" ); -pci_setup_iommu (&s->pci_bus, raven_pcihost_set_iommu, s); +pci_setup_iommu (&s->pci_bus, &raven_iommu_ops, s); h->bus = &s->pci_bus;
hw/pci-host/sabre.c +5 -1
@@ -113,6 +113,10 @@ static AddressSpace *sabre_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &is->iommu_as; } +static const PCIIOMMUOps sabre_iommu_ops = { + .get_address_space = sabre_pci_dma_iommu, +}; +static void sabre_config_write (void *opaque,hwaddr addr,uint64_t val,unsigned size) {
@@ -403,7 +407,7 @@ static void sabre_realize(DeviceState *dev, Error **errp)/* IOMMU */ memory_region_add_subregion_overlap (&s->sabre_config,0x200 ,sysbus_mmio_get_region (SYS_BUS_DEVICE (s->iommu),0 ),1 ); -pci_setup_iommu (phb->bus, sabre_pci_dma_iommu, s->iommu); +pci_setup_iommu (phb->bus, &sabre_iommu_ops, s->iommu);/* APB secondary busses */ pci_dev = pci_new_multifunction(PCI_DEVFN(1, 0), true,
hw/pci/pci.c +13 -5
@@ -2665,7 +2665,13 @@ AddressSpace *pci_device_iommu_address_space(PCIDevice *dev)PCIBus *iommu_bus = bus;uint8_t devfn = dev->devfn; -while (iommu_bus && !iommu_bus->iommu_fn && iommu_bus->parent_dev) { +/* +* get_address_space() callback is mandatory, so needs to ensure its +* presence in the iommu_bus search. +*/ +while (iommu_bus && (!iommu_bus->iommu_ops || + !iommu_bus->iommu_ops->get_address_space) && + iommu_bus->parent_dev) {PCIBus *parent_bus =pci_get_bus (iommu_bus->parent_dev);/*
@@ -2704,15 +2710,17 @@ AddressSpace *pci_device_iommu_address_space(PCIDevice *dev) iommu_bus = parent_bus; } -if (iommu_bus && iommu_bus->iommu_fn) { -return iommu_bus->iommu_fn (bus, iommu_bus->iommu_opaque, devfn); +if (iommu_bus && iommu_bus->iommu_ops && + iommu_bus->iommu_ops->get_address_space) { +return iommu_bus->iommu_ops->get_address_space (bus, + iommu_bus->iommu_opaque, devfn); }return &address_space_memory; } -void pci_setup_iommu (PCIBus *bus,PCIIOMMUFunc fn,void *opaque) +void pci_setup_iommu (PCIBus *bus,const PCIIOMMUOps *ops,void *opaque) { - bus->iommu_fn = fn; + bus->iommu_ops = ops; bus->iommu_opaque = opaque; }
hw/ppc/ppc440_pcix.c +5 -1
@@ -442,6 +442,10 @@ static AddressSpace *ppc440_pcix_set_iommu(PCIBus *b, void *opaque, int devfn)return &s->bm_as; } +static const PCIIOMMUOps ppc440_iommu_ops = { + .get_address_space = ppc440_pcix_set_iommu, +}; +/* The default pci_host_data_{read,write} functions in pci/pci_host.c * deny access to registers without bit 31 set but our clients want * this to work so we have to override these here */
@@ -487,7 +491,7 @@ static void ppc440_pcix_realize(DeviceState *dev, Error **errp)memory_region_init (&s->bm,OBJECT (s),"bm-ppc440-pcix" ,UINT64_MAX );memory_region_add_subregion (&s->bm,0x0 , &s->busmem);address_space_init (&s->bm_as, &s->bm,"pci-bm" ); -pci_setup_iommu (h->bus, ppc440_pcix_set_iommu, s); +pci_setup_iommu (h->bus, &ppc440_iommu_ops, s);memory_region_init (&s->container,OBJECT (s),"pci-container" ,PCI_ALL_SIZE ); memory_region_init_io(&h->conf_mem, OBJECT(s), &pci_host_conf_le_ops,
hw/ppc/spapr_pci.c +5 -1
@@ -771,6 +771,10 @@ static AddressSpace *spapr_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &phb->iommu_as; } +static const PCIIOMMUOps spapr_iommu_ops = { + .get_address_space = spapr_pci_dma_iommu, +}; +static char *spapr_phb_vfio_get_loc_code (SpaprPhbState *sphb,PCIDevice *pdev) {char *path =NULL , *buf =NULL , *host =NULL ;
@@ -1956,7 +1960,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)memory_region_add_subregion (&sphb->iommu_root,SPAPR_PCI_MSI_WINDOW , &sphb->msiwindow); -pci_setup_iommu (bus, spapr_pci_dma_iommu, sphb); +pci_setup_iommu (bus, &spapr_iommu_ops, sphb);pci_bus_set_route_irq_fn (bus, spapr_route_intx_pin_to_irq);
hw/s390x/s390-pci-bus.c +6 -2
@@ -635,6 +635,10 @@ static AddressSpace *s390_pci_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &iommu->as; } +static const PCIIOMMUOps s390_iommu_ops = { + .get_address_space = s390_pci_dma_iommu, +}; +static uint8_t set_ind_atomic (uint64_t ind_loc,uint8_t to_be_set) {uint8_t expected, actual;
@@ -749,7 +753,7 @@ static void s390_pcihost_realize(DeviceState *dev, Error **errp) b =pci_register_root_bus (dev,NULL , s390_pci_set_irq, s390_pci_map_irq,NULL ,get_system_memory (),get_system_io (),0 ,64 ,TYPE_PCI_BUS ); -pci_setup_iommu (b, s390_pci_dma_iommu, s); +pci_setup_iommu (b, &s390_iommu_ops, s); bus =BUS (b);qbus_set_hotplug_handler (bus,OBJECT (dev));
@@ -909,7 +913,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev, pdev =PCI_DEVICE (dev);pci_bridge_map_irq (pb, dev->id, s390_pci_map_irq); -pci_setup_iommu (&pb->sec_bus, s390_pci_dma_iommu, s); +pci_setup_iommu (&pb->sec_bus, &s390_iommu_ops, s);qbus_set_hotplug_handler (BUS (&pb->sec_bus),OBJECT (s));
hw/virtio/virtio-iommu.c +5 -1
@@ -236,6 +236,10 @@ static AddressSpace *virtio_iommu_find_add_as(PCIBus *bus, void *opaque,return &sdev->as; } +static const PCIIOMMUOps virtio_iommu_ops = { + .get_address_space = virtio_iommu_find_add_as, +}; +static int virtio_iommu_attach (VirtIOIOMMU *s,struct virtio_iommu_req_attach *req) {
@@ -789,7 +793,7 @@ static void virtio_iommu_device_realize(DeviceState *dev, Error **errp) s->as_by_busptr =g_hash_table_new_full (NULL ,NULL ,NULL , g_free);if (s->primary_bus) { -pci_setup_iommu (s->primary_bus, virtio_iommu_find_add_as, s); +pci_setup_iommu (s->primary_bus, &virtio_iommu_ops, s); }else {error_setg (errp,"VIRTIO-IOMMU is not attached to any PCI bus!" ); }
include/hw/pci/pci.h +6 -2
@@ -487,10 +487,14 @@ void pci_bus_get_w64_range(PCIBus *bus, Range *range);void pci_device_deassert_intx (PCIDevice *dev); -typedef AddressSpace *(*PCIIOMMUFunc )(PCIBus *,void *,int ); +typedef struct PCIIOMMUOps PCIIOMMUOps ; +struct PCIIOMMUOps { +AddressSpace * (*get_address_space)(PCIBus *bus, +void *opaque,int32_t devfn); +};AddressSpace *pci_device_iommu_address_space (PCIDevice *dev); -void pci_setup_iommu (PCIBus *bus,PCIIOMMUFunc fn,void *opaque); +void pci_setup_iommu (PCIBus *bus,const PCIIOMMUOps *iommu_ops,void *opaque);static inline void pci_set_byte (uint8_t *config,uint8_t val)
include/hw/pci/pci_bus.h +1 -1
@@ -29,7 +29,7 @@ enum PCIBusFlags {struct PCIBus {BusState qbus;enum PCIBusFlags flags; -PCIIOMMUFunc iommu_fn; +const PCIIOMMUOps *iommu_ops;void *iommu_opaque;uint8_t devfn_min;uint32_t slot_reserved_mask;
Signature · 1 line
2.7.4
RFC v10 4/25 hw/pci: introduce pci_device_get_iommu_attr()
In reply to Liu Yi LThis patch adds pci_device_get_iommu_attr() to get vIOMMU attributes. e.g. if nesting IOMMU wanted.
--- hw/pci/pci.c | 35 ++++++++++++++++++++++++++++++----- include/hw/pci/pci.h | 7 +++++++ 2 files changed, 37 insertions(+), 5 deletions(-)
hw/pci/pci.c +30 -5
@@ -2659,7 +2659,8 @@ static void pci_device_class_base_init(ObjectClass *klass, void *data) } } -AddressSpace *pci_device_iommu_address_space (PCIDevice *dev) +static void pci_device_get_iommu_bus_devfn (PCIDevice *dev, +PCIBus **pbus,uint8_t *pdevfn) {PCIBus *bus =pci_get_bus (dev);PCIBus *iommu_bus = bus;
@@ -2710,14 +2711,38 @@ AddressSpace *pci_device_iommu_address_space(PCIDevice *dev) iommu_bus = parent_bus; } -if (iommu_bus && iommu_bus->iommu_ops && - iommu_bus->iommu_ops->get_address_space) { -return iommu_bus->iommu_ops->get_address_space (bus, - iommu_bus->iommu_opaque, devfn); + *pbus = iommu_bus; + *pdevfn = devfn; +} + +AddressSpace *pci_device_iommu_address_space (PCIDevice *dev) +{ +PCIBus *bus; +uint8_t devfn; + +pci_device_get_iommu_bus_devfn (dev, &bus, &devfn); +if (bus && bus->iommu_ops && + bus->iommu_ops->get_address_space) { +return bus->iommu_ops->get_address_space (bus, + bus->iommu_opaque, devfn); }return &address_space_memory; } +int pci_device_get_iommu_attr (PCIDevice *dev,IOMMUAttr attr,void *data) +{ +PCIBus *bus; +uint8_t devfn; + +pci_device_get_iommu_bus_devfn (dev, &bus, &devfn); +if (bus && bus->iommu_ops && + bus->iommu_ops->get_iommu_attr) { +return bus->iommu_ops->get_iommu_attr (bus, bus->iommu_opaque, + devfn, attr, data); + } +return -ENOENT ; +} +void pci_setup_iommu (PCIBus *bus,const PCIIOMMUOps *ops,void *opaque) { bus->iommu_ops = ops;
include/hw/pci/pci.h +7 -0
@@ -487,13 +487,20 @@ void pci_bus_get_w64_range(PCIBus *bus, Range *range);void pci_device_deassert_intx (PCIDevice *dev); +typedef enum IOMMUAttr { +IOMMU_WANT_NESTING , +}IOMMUAttr ; +typedef struct PCIIOMMUOps PCIIOMMUOps ;struct PCIIOMMUOps {AddressSpace * (*get_address_space)(PCIBus *bus,void *opaque,int32_t devfn); +int (*get_iommu_attr)(PCIBus *bus,void *opaque,int32_t devfn, +IOMMUAttr attr,void *data); };AddressSpace *pci_device_iommu_address_space (PCIDevice *dev); +int pci_device_get_iommu_attr (PCIDevice *dev,IOMMUAttr attr,void *data);void pci_setup_iommu (PCIBus *bus,const PCIIOMMUOps *iommu_ops,void *opaque); static inline void
Signature · 1 line
2.7.4
RFC v10 5/25 intel_iommu: add get_iommu_attr() callback
In reply to Liu Yi LReturn vIOMMU attribute to caller. e.g. VFIO call via PCI layer.
--- hw/i386/intel_iommu.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+)
hw/i386/intel_iommu.c +23 -0
@@ -3440,6 +3440,28 @@ VTDAddressSpace *vtd_find_add_as(IntelIOMMUState *s, PCIBus *bus, int devfn)return vtd_dev_as; } +static int vtd_dev_get_iommu_attr (PCIBus *bus,void *opaque,int32_t devfn, +IOMMUAttr attr,void *data) +{ +int ret =0 ; + +assert (0 <= devfn && devfn <PCI_DEVFN_MAX ); + +switch (attr) { +case IOMMU_WANT_NESTING : + { +bool *pdata = data; + +/* return false until vSVA is ready */ + *pdata = false; +break ; + } +default : + ret = -ENOENT ; + } +return ret; +} +static uint64_t get_naturally_aligned_size (uint64_t start,uint64_t size,int gaw) {
@@ -3735,6 +3757,7 @@ static AddressSpace *vtd_host_dma_iommu(PCIBus *bus, void *opaque, int devfn)static PCIIOMMUOps vtd_iommu_ops = { .get_address_space = vtd_host_dma_iommu, + .get_iommu_attr = vtd_dev_get_iommu_attr, };static bool vtd_decide_config (IntelIOMMUState *s,Error **errp)
Signature · 1 line
2.7.4
RFC v10 6/25 vfio: pass nesting requirement into vfio_get_group()
In reply to Liu Yi LThis patch passes the nesting requirement into vfio_get_group() to indicate whether VFIO_TYPE1_NESTING_IOMMU is required.
--- hw/vfio/ap.c | 2 +- hw/vfio/ccw.c | 2 +- hw/vfio/common.c | 3 ++- hw/vfio/pci.c | 9 ++++++++- hw/vfio/platform.c | 2 +- include/hw/vfio/vfio-common.h | 3 ++- 6 files changed, 15 insertions(+), 6 deletions(-)
hw/vfio/ap.c +1 -1
@@ -83,7 +83,7 @@ static VFIOGroup *vfio_ap_get_group(VFIOAPDevice *vapdev, Error **errp)g_free (group_path); -return vfio_get_group (groupid, &address_space_memory, errp); +return vfio_get_group (groupid, &address_space_memory, false, errp); }static void vfio_ap_realize (DeviceState *dev,Error **errp)
hw/vfio/ccw.c +1 -1
@@ -621,7 +621,7 @@ static VFIOGroup *vfio_ccw_get_group(S390CCWDevice *cdev, Error **errp)return NULL ; } -return vfio_get_group (groupid, &address_space_memory, errp); +return vfio_get_group (groupid, &address_space_memory, false, errp); }static void vfio_ccw_realize (DeviceState *dev,Error **errp)
hw/vfio/common.c +2 -1
@@ -1457,7 +1457,8 @@ static void vfio_disconnect_container(VFIOGroup *group) } } -VFIOGroup *vfio_get_group (int groupid,AddressSpace *as,Error **errp) +VFIOGroup *vfio_get_group (int groupid,AddressSpace *as, +bool want_nested,Error **errp) {VFIOGroup *group;char path[32 ];
hw/vfio/pci.c +8 -1
@@ -2711,6 +2711,7 @@ static void vfio_realize(PCIDevice *pdev, Error **errp)int groupid;int i, ret;bool is_mdev; +bool want_nested;if (!vdev->vbasedev.sysfsdev) { if (!(~vdev->host.domain || ~vdev->host.bus ||
@@ -2768,7 +2769,13 @@ static void vfio_realize(PCIDevice *pdev, Error **errp)trace_vfio_realize (vdev->vbasedev.name, groupid); - group =vfio_get_group (groupid,pci_device_iommu_address_space (pdev), errp); +if (pci_device_get_iommu_attr (pdev, +IOMMU_WANT_NESTING , &want_nested)) { + want_nested = false; + } + + group =vfio_get_group (groupid,pci_device_iommu_address_space (pdev), + want_nested, errp);if (!group) { goto error; }
hw/vfio/platform.c +1 -1
@@ -580,7 +580,7 @@ static int vfio_base_device_init(VFIODevice *vbasedev, Error **errp)trace_vfio_platform_base_device_init (vbasedev->name, groupid); - group =vfio_get_group (groupid, &address_space_memory, errp); + group =vfio_get_group (groupid, &address_space_memory, false, errp);if (!group) {return -ENOENT ; }
include/hw/vfio/vfio-common.h +2 -1
@@ -174,7 +174,8 @@ void vfio_region_mmaps_set_enabled(VFIORegion *region, bool enabled);void vfio_region_exit (VFIORegion *region);void vfio_region_finalize (VFIORegion *region);void vfio_reset_handler (void *opaque); -VFIOGroup *vfio_get_group (int groupid,AddressSpace *as,Error **errp); +VFIOGroup *vfio_get_group (int groupid,AddressSpace *as, +bool want_nested,Error **errp);void vfio_put_group (VFIOGroup *group);int vfio_get_device (VFIOGroup *group,const char *name,VFIODevice *vbasedev,Error **errp);
Signature · 1 line
2.7.4
RFC v10 7/25 vfio: check VFIO_TYPE1_NESTING_IOMMU support
In reply to Liu Yi LVFIO needs to check VFIO_TYPE1_NESTING_IOMMU support with Kernel before further using it. e.g. requires to check IOMMU UAPI support.
Referred patch from Eric Auger: https://patchwork.kernel.org/patch/11040499/
--- hw/vfio/common.c | 37 ++++++++++++++++++++++++++----------- 1 file changed, 26 insertions(+), 11 deletions(-)
hw/vfio/common.c +26 -11
@@ -1151,30 +1151,44 @@ static void vfio_put_address_space(VFIOAddressSpace *space) }/* -* vfio_get_iommu_type - selects the richest iommu_type (v2 first) +* vfio_get_iommu_type - selects the richest iommu_type (NESTING first) */ static int vfio_get_iommu_type (VFIOContainer *container, +bool want_nested,Error **errp) { -int iommu_types[] = { VFIO_TYPE1v2_IOMMU,VFIO_TYPE1_IOMMU , +int iommu_types[] = {VFIO_TYPE1_NESTING_IOMMU , + VFIO_TYPE1v2_IOMMU,VFIO_TYPE1_IOMMU , VFIO_SPAPR_TCE_v2_IOMMU,VFIO_SPAPR_TCE_IOMMU }; -int i; +int i, ret = -EINVAL ;for (i =0 ; i <ARRAY_SIZE (iommu_types); i++) {if (ioctl (container->fd,VFIO_CHECK_EXTENSION , iommu_types[i])) { -return iommu_types[i]; +if (iommu_types[i] ==VFIO_TYPE1_NESTING_IOMMU ) { +if (!want_nested) { +continue ; + } + } + ret = iommu_types[i]; +break ; } } -error_setg (errp,"No available IOMMU models" ); -return -EINVAL ; + +if (ret <0 ) { +error_setg (errp,"No available IOMMU models" ); + }else if (want_nested && ret !=VFIO_TYPE1_NESTING_IOMMU ) { +error_setg (errp,"Nested mode requested but not supported" ); + ret = -EINVAL ; + } +return ret; }static int vfio_init_container (VFIOContainer *container,int group_fd, -Error **errp) +bool want_nested,Error **errp) {int iommu_type, ret; - iommu_type =vfio_get_iommu_type (container, errp); + iommu_type =vfio_get_iommu_type (container, want_nested, errp);if (iommu_type <0 ) {return iommu_type; }
@@ -1205,7 +1219,7 @@ static int vfio_init_container(VFIOContainer *container, int group_fd, }static int vfio_connect_container (VFIOGroup *group,AddressSpace *as, -Error **errp) +bool want_nested,Error **errp) {VFIOContainer *container;int ret, fd;
@@ -1276,12 +1290,13 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as,QLIST_INIT (&container->giommu_list);QLIST_INIT (&container->hostwin_list); - ret =vfio_init_container (container, group->fd, errp); + ret =vfio_init_container (container, group->fd, want_nested, errp);if (ret) { goto free_container_exit; }switch (container->iommu_type) { +case VFIO_TYPE1_NESTING_IOMMU :case VFIO_TYPE1v2_IOMMU:case VFIO_TYPE1_IOMMU : {
@@ -1502,7 +1517,7 @@ VFIOGroup *vfio_get_group(int groupid, AddressSpace *as, group->groupid = groupid;QLIST_INIT (&group->device_list); -if (vfio_connect_container (group, as, errp)) { +if (vfio_connect_container (group, as, want_nested, errp)) {error_prepend (errp,"failed to setup container for group %d: " , groupid); goto close_fd_exit;
Signature · 1 line
2.7.4
RFC v10 8/25 hw/iommu: introduce HostIOMMUContext
In reply to Liu Yi LCurrently, many platform vendors provide the capability of dual stage
DMA address translation in hardware. For example, nested translation
on Intel VT-d scalable mode, nested stage translation on ARM SMMUv3,
and etc. In dual stage DMA address translation, there are two stages
address translation, stage-1 (a.k.a first-level) and stage-2 (a.k.a
second-level) translation structures. Stage-1 translation results are
also subjected to stage-2 translation structures. Take vSVA (Virtual
Shared Virtual Addressing) as an example, guest IOMMU driver owns
stage-1 translation structures (covers GVA->GPA translation), and host
IOMMU driver owns stage-2 translation structures (covers GPA->HPA
translation). VMM is responsible to bind stage-1 translation structures
to host, thus hardware could achieve GVA->GPA and then GPA->HPA
translation. For more background on SVA, refer the below links.
- https://www.youtube.com/watch?v=Kq_nfGK5MwQ
- https://events19.lfasiallc.com/wp-content/uploads/2017/11/Shared-Virtual-Memory-in-KVM_Yi-Liu.pdf
In QEMU, vIOMMU emulators expose IOMMUs to VM per their own spec (e.g. Intel VT-d spec). Devices are pass-through to guest via device pass- through components like VFIO. VFIO is a userspace driver framework which exposes host IOMMU programming capability to userspace in a secure manner. e.g. IOVA MAP/UNMAP requests. Information, different from map/unmap notifications need to be passed from QEMU vIOMMU device to/from the host IOMMU driver through the VFIO/IOMMU layer:
1) PASID allocation (allow host to intercept in PASID allocation)
2) bind stage-1 translation structures to host
3) propagate stage-1 cache invalidation to host
4) DMA address translation fault (I/O page fault) servicing etc.
With the above new interactions in QEMU, it requires an abstract layer to facilitate the above operations and expose to vIOMMU emulators as an explicit way for vIOMMU emulators call into VFIO. This patch introduces HostIOMMUContext to serve it. The HostIOMMUContext is an object which allows to manage the stage-1 translation when a vIOMMU is implemented upon physical IOMMU nested paging (VFIO case). It is an abstract object which needs to be derived for each vIOMMU immplementation based on physical nested paging. An HostIOMMUContext derived object will be passed to each VFIO device protected by a vIOMMU using physical nested paging.
This patchg also introduces HostIOMMUContextClass to provide methods for vIOMMU emulators to propagate dual-stage translation related requests to host. As a beginning, PASID allocation/free are defined to propagate PASID allocation/free requests to host which is required for the vendors which manage PASID in system-wide.
---
rfcv9 -> rfcv10:
*) adjust to meson build
--- hw/Kconfig | 3 ++ hw/iommu/Kconfig | 4 ++ hw/iommu/host_iommu_context.c | 97 +++++++++++++++++++++++++++++++++++ hw/iommu/meson.build | 6 +++ hw/meson.build | 1 + include/hw/iommu/host_iommu_context.h | 75 +++++++++++++++++++++++++++ 6 files changed, 186 insertions(+) create mode 100644 hw/iommu/Kconfig create mode 100644 hw/iommu/host_iommu_context.c create mode 100644 hw/iommu/meson.build create mode 100644 include/hw/iommu/host_iommu_context.h
hw/Kconfig +3 -0
@@ -65,6 +65,9 @@ source tricore/Kconfig source unicore32/Kconfig source xtensa/Kconfig +# iommu Kconfig +source iommu/Kconfig + # Symbols used by multiple targets config TEST_DEVICES bool
hw/iommu/Kconfig new file +4 -0
@@ -0,0 +1,4 @@ +config IOMMU + bool + default y + depends on LINUX
hw/iommu/host_iommu_context.c new file +97 -0
97 lines not on this page · raw message
hw/iommu/meson.build new file +6 -0
@@ -0,0 +1,6 @@ +iommu_ss = ss.source_set() +iommu_ss.add(files( + 'host_iommu_context.c', +)) + +specific_ss.add_all(when: 'CONFIG_IOMMU', if_true: iommu_ss)
hw/meson.build +1 -0
@@ -65,3 +65,4 @@ subdir('sparc64') subdir('tricore') subdir('unicore32') subdir('xtensa') +subdir('iommu')
include/hw/iommu/host_iommu_context.h new file +75 -0
75 lines not on this page · raw message
Signature · 1 line
2.7.4
RFC v10 9/25 hw/pci: introduce pci_device_set/unset_iommu_context()
In reply to Liu Yi LFor nesting IOMMU translation capable platforms, vIOMMUs running on such system could be implemented upon physical IOMMU nested paging (VFIO case). vIOMMU advertises such implementation by "want_nested" attribute to PCIe devices (e.g. VFIO PCI). Once "want_nested" is satisfied, device (VFIO case) should set HostIOMMUContext to vIOMMU, thus vIOMMU could manage stage-1 translation. DMAs out from such devices would be protected through the stage-1 page tables owned by guest together with stage-2 page tables owned by host.
This patch adds pci_device_set/unset_iommu_context() to set/unset HostIOMMUContext for a given PCIe device (VFIO case). Caller of set should fail if set operation failed.
---
rfcv5 (v2) -> rfcv6:
*) pci_device_set_iommu_context() returns 0 if callback is not implemented.
--- hw/pci/pci.c | 28 ++++++++++++++++++++++++++++ include/hw/pci/pci.h | 10 ++++++++++ 2 files changed, 38 insertions(+)
hw/pci/pci.c +28 -0
@@ -2743,6 +2743,34 @@ int pci_device_get_iommu_attr(PCIDevice *dev, IOMMUAttr attr, void *data)return -ENOENT ; } +int pci_device_set_iommu_context (PCIDevice *dev, +HostIOMMUContext *iommu_ctx) +{ +PCIBus *bus; +uint8_t devfn; + +pci_device_get_iommu_bus_devfn (dev, &bus, &devfn); +if (bus && bus->iommu_ops && + bus->iommu_ops->set_iommu_context) { +return bus->iommu_ops->set_iommu_context (bus, + bus->iommu_opaque, devfn, iommu_ctx); + } +return 0 ; +} + +void pci_device_unset_iommu_context (PCIDevice *dev) +{ +PCIBus *bus; +uint8_t devfn; + +pci_device_get_iommu_bus_devfn (dev, &bus, &devfn); +if (bus && bus->iommu_ops && + bus->iommu_ops->unset_iommu_context) { + bus->iommu_ops->unset_iommu_context (bus, + bus->iommu_opaque, devfn); + } +} +void pci_setup_iommu (PCIBus *bus,const PCIIOMMUOps *ops,void *opaque) { bus->iommu_ops = ops;
include/hw/pci/pci.h +10 -0
@@ -9,6 +9,8 @@#include "hw/pci/pcie.h" +#include "hw/iommu/host_iommu_context.h" +extern bool pci_available;/* PCI bus */
@@ -497,10 +499,18 @@ struct PCIIOMMUOps { void *opaque, int32_t devfn);int (*get_iommu_attr)(PCIBus *bus,void *opaque,int32_t devfn,IOMMUAttr attr,void *data); +int (*set_iommu_context)(PCIBus *bus,void *opaque, +int32_t devfn, +HostIOMMUContext *iommu_ctx); +void (*unset_iommu_context)(PCIBus *bus,void *opaque, +int32_t devfn); };AddressSpace *pci_device_iommu_address_space (PCIDevice *dev);int pci_device_get_iommu_attr (PCIDevice *dev,IOMMUAttr attr,void *data); +int pci_device_set_iommu_context (PCIDevice *dev, +HostIOMMUContext *iommu_ctx); +void pci_device_unset_iommu_context (PCIDevice *dev);void pci_setup_iommu (PCIBus *bus,const PCIIOMMUOps *iommu_ops,void *opaque); static inline void
Signature · 1 line
2.7.4
RFC v10 10/25 intel_iommu: add set/unset_iommu_context callback
In reply to Liu Yi LThis patch adds set/unset_iommu_context() impelementation in Intel vIOMMU. PCIe devices (VFIO case) sets HostIOMMUContext to vIOMMU as an ack of vIOMMU's "want_nested" attribute. Thus vIOMMU could build DMA protection based on nested paging of host IOMMU.
--- hw/i386/intel_iommu.c | 71 ++++++++++++++++++++++++++++++++++++++++--- include/hw/i386/intel_iommu.h | 21 ++++++++++--- 2 files changed, 83 insertions(+), 9 deletions(-)
hw/i386/intel_iommu.c +66 -5
@@ -3358,23 +3358,33 @@ static const MemoryRegionOps vtd_mem_ir_ops = { }, }; -VTDAddressSpace *vtd_find_add_as (IntelIOMMUState *s,PCIBus *bus,int devfn) +/** +* Fetch a VTDBus instance for given PCIBus. If no existing instance, +* allocate one. +*/ +static VTDBus *vtd_find_add_bus (IntelIOMMUState *s,PCIBus *bus) {uintptr_t key = (uintptr_t )bus;VTDBus *vtd_bus =g_hash_table_lookup (s->vtd_as_by_busptr, &key); -VTDAddressSpace *vtd_dev_as; -char name[128 ];if (!vtd_bus) {uintptr_t *new_key =g_malloc (sizeof (*new_key)); *new_key = (uintptr_t )bus;/* No corresponding free() */ - vtd_bus =g_malloc0 (sizeof (VTDBus) +sizeof (VTDAddressSpace *) * \ -PCI_DEVFN_MAX ); + vtd_bus =g_malloc0 (sizeof (VTDBus)); vtd_bus->bus = bus;g_hash_table_insert (s->vtd_as_by_busptr, new_key, vtd_bus); } +return vtd_bus; +} +VTDAddressSpace *vtd_find_add_as (IntelIOMMUState *s,PCIBus *bus,int devfn) +{ +VTDBus *vtd_bus; +VTDAddressSpace *vtd_dev_as; +char name[128 ]; + + vtd_bus =vtd_find_add_bus (s, bus); vtd_dev_as = vtd_bus->dev_as[devfn];if (!vtd_dev_as) {
@@ -3462,6 +3472,55 @@ static int vtd_dev_get_iommu_attr(PCIBus *bus, void *opaque, int32_t devfn,return ret; } +static int vtd_dev_set_iommu_context (PCIBus *bus,void *opaque, +int devfn, +HostIOMMUContext *iommu_ctx) +{ +IntelIOMMUState *s = opaque; +VTDBus *vtd_bus; +VTDHostIOMMUContext *vtd_dev_icx; + +assert (0 <= devfn && devfn <PCI_DEVFN_MAX ); + + vtd_bus =vtd_find_add_bus (s, bus); + +vtd_iommu_lock (s); + + vtd_dev_icx = vtd_bus->dev_icx[devfn]; + +assert (!vtd_dev_icx); + + vtd_bus->dev_icx[devfn] = vtd_dev_icx = +g_malloc0 (sizeof (VTDHostIOMMUContext)); + vtd_dev_icx->vtd_bus = vtd_bus; + vtd_dev_icx->devfn = (uint8_t )devfn; + vtd_dev_icx->iommu_state = s; + vtd_dev_icx->iommu_ctx = iommu_ctx; + +vtd_iommu_unlock (s); + +return 0 ; +} + +static void vtd_dev_unset_iommu_context (PCIBus *bus,void *opaque,int devfn) +{ +IntelIOMMUState *s = opaque; +VTDBus *vtd_bus; +VTDHostIOMMUContext *vtd_dev_icx; + +assert (0 <= devfn && devfn <PCI_DEVFN_MAX ); + + vtd_bus =vtd_find_add_bus (s, bus); + +vtd_iommu_lock (s); + + vtd_dev_icx = vtd_bus->dev_icx[devfn]; +g_free (vtd_dev_icx); + vtd_bus->dev_icx[devfn] =NULL ; + +vtd_iommu_unlock (s); +} +static uint64_t get_naturally_aligned_size (uint64_t start,uint64_t size,int gaw) {
@@ -3758,6 +3817,8 @@ static AddressSpace *vtd_host_dma_iommu(PCIBus *bus, void *opaque, int devfn)static PCIIOMMUOps vtd_iommu_ops = { .get_address_space = vtd_host_dma_iommu, .get_iommu_attr = vtd_dev_get_iommu_attr, + .set_iommu_context = vtd_dev_set_iommu_context, + .unset_iommu_context = vtd_dev_unset_iommu_context, };static bool vtd_decide_config (IntelIOMMUState *s,Error **errp)
include/hw/i386/intel_iommu.h +17 -4
@@ -64,6 +64,7 @@ typedef union VTD_IR_TableEntry VTD_IR_TableEntry;typedef union VTD_IR_MSIAddress VTD_IR_MSIAddress ;typedef struct VTDPASIDDirEntry VTDPASIDDirEntry ;typedef struct VTDPASIDEntry VTDPASIDEntry ; +typedef struct VTDHostIOMMUContext VTDHostIOMMUContext ;/* Context-Entry */ struct VTDContextEntry {
@@ -112,10 +113,20 @@ struct VTDAddressSpace {IOVATree *iova_tree;/* Traces mapped IOVA ranges */ }; +struct VTDHostIOMMUContext { +VTDBus *vtd_bus; +uint8_t devfn; +HostIOMMUContext *iommu_ctx; +IntelIOMMUState *iommu_state; +}; +struct VTDBus { -PCIBus * bus;/* A reference to the bus to provide translation for */ +/* A reference to the bus to provide translation for */ +PCIBus *bus;/* A table of VTDAddressSpace objects indexed by devfn */ -VTDAddressSpace *dev_as[]; +VTDAddressSpace *dev_as[PCI_DEVFN_MAX ]; +/* A table of VTDHostIOMMUContext objects indexed by devfn */ +VTDHostIOMMUContext *dev_icx[PCI_DEVFN_MAX ]; };struct VTDIOTLBEntry {
@@ -269,8 +280,10 @@ struct IntelIOMMUState {bool dma_drain;/* Whether DMA r/w draining enabled */ /* -* Protects IOMMU states in general. Currently it protects the -* per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace. +* iommu_lock protects below: +* - per-IOMMU IOTLB caches +* - context entry cache in VTDAddressSpace +* - HostIOMMUContext pointer cached in vIOMMU */ QemuMutex iommu_lock; };
Signature · 1 line
2.7.4
RFC v10 11/25 vfio/common: provide PASID alloc/free hooks
In reply to Liu Yi LThis patch defines vfio_host_iommu_context_info, implements the PASID alloc/free hooks defined in HostIOMMUContextClass.
--- hw/vfio/common.c | 66 +++++++++++++++++++++++++++++++++++ include/hw/iommu/host_iommu_context.h | 3 ++ include/hw/vfio/vfio-common.h | 4 +++ 3 files changed, 73 insertions(+)
hw/vfio/common.c +66 -0
@@ -1183,6 +1183,50 @@ static int vfio_get_iommu_type(VFIOContainer *container,return ret; } +static int vfio_host_iommu_ctx_pasid_alloc (HostIOMMUContext *iommu_ctx, +uint32_t min,uint32_t max, +uint32_t *pasid) +{ +VFIOContainer *container =container_of (iommu_ctx, + VFIOContainer, iommu_ctx); +struct vfio_iommu_type1_pasid_request req; +int ret =0 ; + + req.argsz =sizeof (req); + req.flags =VFIO_IOMMU_FLAG_ALLOC_PASID ; + req.range.min = min; + req.range.max = max; + + ret =ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req); +if (ret <0 ) { +error_report ("%s: alloc failed (%m)" , __func__); +return ret; + } + *pasid = ret; +return 0 ; +} + +static int vfio_host_iommu_ctx_pasid_free (HostIOMMUContext *iommu_ctx, +uint32_t pasid) +{ +VFIOContainer *container =container_of (iommu_ctx, + VFIOContainer, iommu_ctx); +struct vfio_iommu_type1_pasid_request req; + +int ret =0 ; + + req.argsz =sizeof (req); + req.flags =VFIO_IOMMU_FLAG_FREE_PASID ; + req.range.min = pasid; + req.range.max = pasid +1 ; + + ret =ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req); +if (ret) { +error_report ("%s: free failed (%m)" , __func__); + } +return ret; +} +static int vfio_init_container (VFIOContainer *container,int group_fd,bool want_nested,Error **errp) {
@@ -1802,3 +1846,25 @@ int vfio_eeh_as_op(AddressSpace *as, uint32_t op) }return vfio_eeh_container_op (container, op); } + +static void vfio_host_iommu_context_class_init (ObjectClass *klass, +void *data) +{ +HostIOMMUContextClass *hicxc =HOST_IOMMU_CONTEXT_CLASS (klass); + + hicxc->pasid_alloc = vfio_host_iommu_ctx_pasid_alloc; + hicxc->pasid_free = vfio_host_iommu_ctx_pasid_free; +} + +static const TypeInfo vfio_host_iommu_context_info = { + .parent =TYPE_HOST_IOMMU_CONTEXT , + .name =TYPE_VFIO_HOST_IOMMU_CONTEXT , + .class_init = vfio_host_iommu_context_class_init, +}; + +static void vfio_register_types (void ) +{ +type_register_static (&vfio_host_iommu_context_info); +} + +type_init (vfio_register_types)
include/hw/iommu/host_iommu_context.h +3 -0
@@ -33,6 +33,9 @@#define TYPE_HOST_IOMMU_CONTEXT "qemu:host-iommu-context"#define HOST_IOMMU_CONTEXT (obj) \ OBJECT_CHECK(HostIOMMUContext, (obj), TYPE_HOST_IOMMU_CONTEXT) +#define HOST_IOMMU_CONTEXT_CLASS (klass) \ + OBJECT_CLASS_CHECK(HostIOMMUContextClass, (klass), \ + TYPE_HOST_IOMMU_CONTEXT)#define HOST_IOMMU_CONTEXT_GET_CLASS (obj) \ OBJECT_GET_CLASS(HostIOMMUContextClass, (obj), \ TYPE_HOST_IOMMU_CONTEXT)
include/hw/vfio/vfio-common.h +4 -0
@@ -26,12 +26,15 @@#include "qemu/notify.h" #include "ui/console.h" #include "hw/display/ramfb.h" +#include "hw/iommu/host_iommu_context.h" #ifdef CONFIG_LINUX #include <linux/vfio.h> #endif #define VFIO_MSG_PREFIX "vfio %s: " +#define TYPE_VFIO_HOST_IOMMU_CONTEXT "qemu:vfio-host-iommu-context" +enum {VFIO_DEVICE_TYPE_PCI =0 ,VFIO_DEVICE_TYPE_PLATFORM =1 ,
@@ -71,6 +74,7 @@ typedef struct VFIOContainer {MemoryListener listener;MemoryListener prereg_listener;unsigned iommu_type; +HostIOMMUContext iommu_ctx;Error *error;bool initialized;unsigned long pgsizes;
Signature · 1 line
2.7.4
RFC v10 12/25 vfio: init HostIOMMUContext per-container
In reply to Liu Yi LIn this patch, QEMU firstly gets iommu info from kernel to check the supported capabilities by a VFIO_IOMMU_TYPE1_NESTING iommu. And inits HostIOMMUContet instance.
For vfio-pci devices, it could use pci_device_set/unset_iommu() to expose host iommu context to vIOMMU emulators. vIOMMU emulators could make use the methods provided by host iommu context. e.g. propagate requests to host iommu.
--- hw/vfio/common.c | 113 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ hw/vfio/pci.c | 17 +++++++++ 2 files changed, 130 insertions(+)
hw/vfio/common.c +113 -0
@@ -1227,10 +1227,102 @@ static int vfio_host_iommu_ctx_pasid_free(HostIOMMUContext *iommu_ctx,return ret; } +/** +* Get iommu info from host. Caller of this funcion should free +* the memory pointed by the returned pointer stored in @info +* after a successful calling when finished its usage. +*/ +static int vfio_get_iommu_info (VFIOContainer *container, +struct vfio_iommu_type1_info **info) +{ + +size_t argsz =sizeof (struct vfio_iommu_type1_info ); + + *info =g_malloc0 (argsz); + +retry: + (*info)->argsz = argsz; + +if (ioctl (container->fd,VFIO_IOMMU_GET_INFO , *info)) { +g_free (*info); + *info =NULL ; +return -errno; + } + +if (((*info)->argsz > argsz)) { + argsz = (*info)->argsz; + *info =g_realloc (*info, argsz); + goto retry; + } + +return 0 ; +} + +static struct vfio_info_cap_header * +vfio_get_iommu_info_cap (struct vfio_iommu_type1_info *info,uint16_t id) +{ +struct vfio_info_cap_header *hdr; +void *ptr = info; + +if (!(info->flags &VFIO_IOMMU_INFO_CAPS )) { +return NULL ; + } + +for (hdr = ptr + info->cap_offset; hdr != ptr; hdr = ptr + hdr->next) { +if (hdr->id == id) { +return hdr; + } + } + +return NULL ; +} + +static int vfio_get_nesting_iommu_cap (VFIOContainer *container, +struct vfio_iommu_type1_info_cap_nesting **cap_nesting) +{ +struct vfio_iommu_type1_info *info; +struct vfio_info_cap_header *hdr; +struct vfio_iommu_type1_info_cap_nesting *cap; +struct iommu_nesting_info *nest_info; +int ret; +uint32_t minsz, cap_size; + + ret =vfio_get_iommu_info (container, &info); +if (ret) { +return ret; + } + + hdr =vfio_get_iommu_info_cap (info, +VFIO_IOMMU_TYPE1_INFO_CAP_NESTING ); +if (!hdr) { +g_free (info); +return -EINVAL ; + } + + cap =container_of (hdr, + struct vfio_iommu_type1_info_cap_nesting, header); + + nest_info = &cap->info; + minsz = offsetof(struct iommu_nesting_info , vendor); +if (nest_info->argsz < minsz) { +g_free (info); +return -EINVAL ; + } + + cap_size = offsetof(struct vfio_iommu_type1_info_cap_nesting , info) + + nest_info->argsz; + *cap_nesting =g_malloc0 (cap_size); +memcpy (*cap_nesting, cap, cap_size); + +g_free (info); +return 0 ; +} +static int vfio_init_container (VFIOContainer *container,int group_fd,bool want_nested,Error **errp) {int iommu_type, ret; +uint64_t flags =0 ; iommu_type =vfio_get_iommu_type (container, want_nested, errp);if (iommu_type <0 ) {
@@ -1258,6 +1350,27 @@ static int vfio_init_container(VFIOContainer *container, int group_fd,return -errno; } +if (iommu_type ==VFIO_TYPE1_NESTING_IOMMU ) { +struct vfio_iommu_type1_info_cap_nesting *nesting =NULL ; +struct iommu_nesting_info *nest_info; + + ret =vfio_get_nesting_iommu_cap (container, &nesting); +if (ret) { +error_setg_errno (errp, -ret, +"Failed to get nesting iommu cap" ); +return ret; + } + + nest_info = (struct iommu_nesting_info *) &nesting->info; + flags |= (nest_info->features &IOMMU_NESTING_FEAT_SYSWIDE_PASID ) ? +HOST_IOMMU_PASID_REQUEST :0 ; +host_iommu_ctx_init (&container->iommu_ctx, +sizeof (container->iommu_ctx), +TYPE_VFIO_HOST_IOMMU_CONTEXT , + flags); +g_free (nesting); + } + container->iommu_type = iommu_type;return 0 ; }
hw/vfio/pci.c +17 -0
@@ -2704,6 +2704,7 @@ static void vfio_realize(PCIDevice *pdev, Error **errp)VFIOPCIDevice *vdev =PCI_VFIO (pdev);VFIODevice *vbasedev_iter;VFIOGroup *group; +VFIOContainer *container;char *tmp, *subsys, group_path[PATH_MAX ], *group_name;Error *err =NULL ;ssize_t len;
@@ -2780,6 +2781,15 @@ static void vfio_realize(PCIDevice *pdev, Error **errp) goto error; } + container = group->container; +if (container->iommu_ctx.initialized && +pci_device_set_iommu_context (pdev, &container->iommu_ctx)) { +error_setg (errp,"device attachment is denied by vIOMMU, " +"please check host IOMMU nesting capability" ); +vfio_put_group (group); + goto error; + } + QLIST_FOREACH(vbasedev_iter, &group->device_list, next) {if (strcmp (vbasedev_iter->name, vdev->vbasedev.name) ==0 ) {error_setg (errp,"device is already attached" );
@@ -3065,9 +3075,16 @@ static void vfio_instance_finalize(Object *obj)static void vfio_exitfn (PCIDevice *pdev) {VFIOPCIDevice *vdev =PCI_VFIO (pdev); +VFIOContainer *container;vfio_unregister_req_notifier (vdev);vfio_unregister_err_notifier (vdev); + + container = vdev->vbasedev.group->container; +if (container->iommu_ctx.initialized) { +pci_device_unset_iommu_context (pdev); + } +pci_device_set_intx_routing_notifier (&vdev->pdev,NULL );if (vdev->irqchip_change_notifier.notify) {kvm_irqchip_remove_change_notifier (&vdev->irqchip_change_notifier);
Signature · 1 line
2.7.4
RFC v10 13/25 intel_iommu: add virtual command capability support
In reply to Liu Yi LThis patch adds virtual command support to Intel vIOMMU per Intel VT-d 3.1 spec. And adds two virtual commands: allocate pasid and free pasid.
--- hw/i386/intel_iommu.c | 154 ++++++++++++++++++++++++++++++++++++++++- hw/i386/intel_iommu_internal.h | 37 ++++++++++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 10 ++- 4 files changed, 200 insertions(+), 2 deletions(-)
hw/i386/intel_iommu.c +153 -1
@@ -2656,6 +2656,129 @@ static void vtd_handle_iectl_write(IntelIOMMUState *s) } } +static int vtd_request_pasid_alloc (IntelIOMMUState *s,uint32_t *pasid) +{ +VTDHostIOMMUContext *vtd_dev_icx; +int ret =-1 ; + +vtd_iommu_lock (s); + QLIST_FOREACH(vtd_dev_icx, &s->vtd_dev_icx_list, next) { +HostIOMMUContext *iommu_ctx = vtd_dev_icx->iommu_ctx; + +/* +* We'll return the first valid result we got. It's +* a bit hackish in that we don't have a good global +* interface yet to talk to modules like vfio to deliver +* this allocation request, so we're leveraging this +* per-device iommu context to do the same thing just +* to make sure the allocation happens only once. +*/ + ret =host_iommu_ctx_pasid_alloc (iommu_ctx,VTD_HPASID_MIN , +VTD_HPASID_MAX , pasid); +if (!ret) { +break ; + } + } +vtd_iommu_unlock (s); + +return ret; +} + +static int vtd_request_pasid_free (IntelIOMMUState *s,uint32_t pasid) +{ +VTDHostIOMMUContext *vtd_dev_icx; +int ret =-1 ; + +vtd_iommu_lock (s); + QLIST_FOREACH(vtd_dev_icx, &s->vtd_dev_icx_list, next) { +HostIOMMUContext *iommu_ctx = vtd_dev_icx->iommu_ctx; + +/* +* Similar with pasid allocation. We'll free the pasid +* on the first successful free operation. It's a bit +* hackish in that we don't have a good global interface +* yet to talk to modules like vfio to deliver this pasid +* free request, so we're leveraging this per-device iommu +* context to do the same thing just to make sure the free +* happens only once. +*/ + ret =host_iommu_ctx_pasid_free (iommu_ctx, pasid); +if (!ret) { +break ; + } + } +vtd_iommu_unlock (s); + +return ret; +} + +/* +* If IP is not set, set it then return. +* If IP is already set, return. +*/ +static void vtd_vcmd_set_ip (IntelIOMMUState *s) +{ + s->vcrsp =1 ; +vtd_set_quad_raw (s,DMAR_VCRSP_REG , + ((uint64_t ) s->vcrsp)); +} + +static void vtd_vcmd_clear_ip (IntelIOMMUState *s) +{ + s->vcrsp &= (~((uint64_t )(0x1 ))); +vtd_set_quad_raw (s,DMAR_VCRSP_REG , + ((uint64_t ) s->vcrsp)); +} + +/* Handle write to Virtual Command Register */ +static int vtd_handle_vcmd_write (IntelIOMMUState *s,uint64_t val) +{ +uint32_t pasid; +int ret =-1 ; + +trace_vtd_reg_write_vcmd (s->vcrsp, val); + +if (!(s->vccap &VTD_VCCAP_PAS ) || + (s->vcrsp &1 )) { +return -1 ; + } + +/* +* Since vCPU should be blocked when the guest VMCD +* write was trapped to here. Should be no other vCPUs +* try to access VCMD if guest software is well written. +* However, we still emulate the IP bit here in case of +* bad guest software. Also align with the spec. +*/ +vtd_vcmd_set_ip (s); + +switch (val &VTD_VCMD_CMD_MASK ) { +case VTD_VCMD_ALLOC_PASID : + ret =vtd_request_pasid_alloc (s, &pasid); +if (ret) { + s->vcrsp |=VTD_VCRSP_SC (VTD_VCMD_NO_AVAILABLE_PASID ); + }else { + s->vcrsp |=VTD_VCRSP_RSLT (pasid); + } +break ; + +case VTD_VCMD_FREE_PASID : + pasid =VTD_VCMD_PASID_VALUE (val); + ret =vtd_request_pasid_free (s, pasid); +if (ret <0 ) { + s->vcrsp |=VTD_VCRSP_SC (VTD_VCMD_FREE_INVALID_PASID ); + } +break ; + +default : + s->vcrsp |=VTD_VCRSP_SC (VTD_VCMD_UNDEFINED_CMD ); +error_report_once ("Virtual Command: unsupported command!!!" ); +break ; + } +vtd_vcmd_clear_ip (s); +return 0 ; +} +static uint64_t vtd_mem_read (void *opaque,hwaddr addr,unsigned size) {IntelIOMMUState *s = opaque;
@@ -2944,6 +3067,23 @@ static void vtd_mem_write(void *opaque, hwaddr addr,vtd_set_long (s, addr, val);break ; +case DMAR_VCMD_REG : +if (!vtd_handle_vcmd_write (s, val)) { +if (size ==4 ) { +vtd_set_long (s, addr, val); + }else { +vtd_set_quad (s, addr, val); + } + } +break ; + +case DMAR_VCMD_REG_HI : +assert (size ==4 ); +if (!vtd_handle_vcmd_write (s, val)) { +vtd_set_long (s, addr, val); + } +break ; +default :if (size ==4 ) {vtd_set_long (s, addr, val);
@@ -3496,6 +3636,7 @@ static int vtd_dev_set_iommu_context(PCIBus *bus, void *opaque, vtd_dev_icx->devfn = (uint8_t )devfn; vtd_dev_icx->iommu_state = s; vtd_dev_icx->iommu_ctx = iommu_ctx; +QLIST_INSERT_HEAD (&s->vtd_dev_icx_list, vtd_dev_icx, next);vtd_iommu_unlock (s);
@@ -3515,7 +3656,10 @@ static void vtd_dev_unset_iommu_context(PCIBus *bus, void *opaque, int devfn)vtd_iommu_lock (s); vtd_dev_icx = vtd_bus->dev_icx[devfn]; -g_free (vtd_dev_icx); +if (vtd_dev_icx) { +QLIST_REMOVE (vtd_dev_icx, next); +g_free (vtd_dev_icx); + } vtd_bus->dev_icx[devfn] =NULL ;vtd_iommu_unlock (s);
@@ -3790,6 +3934,13 @@ static void vtd_init(IntelIOMMUState *s)* Interrupt remapping registers. */ vtd_define_quad (s,DMAR_IRTA_REG ,0 ,0xfffffffffffff80fULL ,0 ); + +/* +* Virtual Command Definitions +*/ +vtd_define_quad (s,DMAR_VCCAP_REG , s->vccap,0 ,0 ); +vtd_define_quad (s,DMAR_VCMD_REG ,0 ,0xffffffffffffffffULL ,0 ); +vtd_define_quad (s,DMAR_VCRSP_REG ,0 ,0 ,0 ); }/* Should not reset address_spaces when reset because devices will still use
@@ -3905,6 +4056,7 @@ static void vtd_realize(DeviceState *dev, Error **errp) }QLIST_INIT (&s->vtd_as_with_notifiers); +QLIST_INIT (&s->vtd_dev_icx_list);qemu_mutex_init (&s->iommu_lock);memset (s->vtd_as_by_bus_num,0 ,sizeof (s->vtd_as_by_bus_num)); memory_region_init_io(&s->csrmem, OBJECT(s), &vtd_mem_ops, s,
hw/i386/intel_iommu_internal.h +37 -0
@@ -85,6 +85,12 @@#define DMAR_MTRRCAP_REG_HI 0x104#define DMAR_MTRRDEF_REG 0x108/* MTRR default type */ #define DMAR_MTRRDEF_REG_HI 0x10c +#define DMAR_VCCAP_REG 0xE00/* Virtual Command Capability Register */ +#define DMAR_VCCAP_REG_HI 0xE04 +#define DMAR_VCMD_REG 0xE10/* Virtual Command Register */ +#define DMAR_VCMD_REG_HI 0xE14 +#define DMAR_VCRSP_REG 0xE20/* Virtual Command Reponse Register */ +#define DMAR_VCRSP_REG_HI 0xE24/* IOTLB registers */ #define DMAR_IOTLB_REG_OFFSET 0xf0/* Offset to the IOTLB registers */
@@ -313,6 +319,37 @@ typedef enum VTDFaultReason {#define VTD_CONTEXT_CACHE_GEN_MAX 0xffffffffUL +/* VCCAP_REG */ +#define VTD_VCCAP_PAS (1UL << 0) + +/* +* The basic idea is to let hypervisor to set a range for available +* PASIDs for VMs. One of the reasons is PASID #0 is reserved by +* RID_PASID usage. We have no idea how many reserved PASIDs in future, +* so here just an evaluated value. Honestly, set it as "1" is enough +* at current stage. +*/ +#define VTD_HPASID_MIN 1 +#define VTD_HPASID_MAX 0xFFFFF + +/* Virtual Command Register */ +enum { +VTD_VCMD_NULL_CMD =0 , +VTD_VCMD_ALLOC_PASID =1 , +VTD_VCMD_FREE_PASID =2 , +VTD_VCMD_CMD_NUM , +}; + +#define VTD_VCMD_CMD_MASK 0xffUL +#define VTD_VCMD_PASID_VALUE (val) (((val) >> 8) & 0xfffff) + +#define VTD_VCRSP_RSLT (val) ((val) << 8) +#define VTD_VCRSP_SC (val) (((val) & 0x3) << 1) + +#define VTD_VCMD_UNDEFINED_CMD 1ULL +#define VTD_VCMD_NO_AVAILABLE_PASID 2ULL +#define VTD_VCMD_FREE_INVALID_PASID 2ULL +/* Interrupt Entry Cache Invalidation Descriptor: VT-d 6.5.2.7. */ struct VTDInvDescIEC {uint32_t type:4 ;/* Should always be 0x4 */
hw/i386/trace-events +1 -0
@@ -51,6 +51,7 @@ vtd_reg_write_gcmd(uint32_t status, uint32_t val) "status 0x%"PRIx32" value 0x%" vtd_reg_write_fectl(uint32_t value) "value 0x%"PRIx32 vtd_reg_write_iectl(uint32_t value) "value 0x%"PRIx32 vtd_reg_ics_clear_ip(void) "" +vtd_reg_write_vcmd(uint32_t status, uint32_t val) "status 0x%"PRIx32" value 0x%"PRIx32 vtd_dmar_translate(uint8_t bus, uint8_t slot, uint8_t func, uint64_t iova, uint64_t gpa, uint64_t mask) "dev %02x:%02x.%02x iova 0x%"PRIx64" -> gpa 0x%"PRIx64" mask 0x%"PRIx64 vtd_dmar_enable(bool en) "enable %d" vtd_dmar_fault(uint16_t sid, int fault, uint64_t addr, bool is_write) "sid 0x%"PRIx16" fault %d addr 0x%"PRIx64" write %d"
include/hw/i386/intel_iommu.h +9 -1
@@ -42,7 +42,7 @@#define VTD_SID_TO_BUS (sid) (((sid) >> 8) & 0xff)#define VTD_SID_TO_DEVFN (sid) ((sid) & 0xff) -#define DMAR_REG_SIZE 0x230 +#define DMAR_REG_SIZE 0xF00#define VTD_HOST_AW_39BIT 39#define VTD_HOST_AW_48BIT 48#define VTD_HOST_ADDRESS_WIDTH VTD_HOST_AW_39BIT
@@ -118,6 +118,7 @@ struct VTDHostIOMMUContext {uint8_t devfn;HostIOMMUContext *iommu_ctx;IntelIOMMUState *iommu_state; +QLIST_ENTRY (VTDHostIOMMUContext) next; };struct VTDBus {
@@ -269,6 +270,9 @@ struct IntelIOMMUState {/* list of registered notifiers */ QLIST_HEAD(, VTDAddressSpace) vtd_as_with_notifiers; +/* list of VTDHostIOMMUContexts */ + QLIST_HEAD(, VTDHostIOMMUContext) vtd_dev_icx_list; +/* interrupt remapping */ bool intr_enabled;/* Whether guest enabled IR */ dma_addr_t intr_root;/* Interrupt remapping table pointer */
@@ -279,6 +283,10 @@ struct IntelIOMMUState {uint8_t aw_bits;/* Host/IOVA address width (in bits) */ bool dma_drain;/* Whether DMA r/w draining enabled */ +/* Virtual Command Register */ +uint64_t vccap;/* The value of vcmd capability reg */ +uint64_t vcrsp;/* Current value of VCMD RSP REG */ +/* * iommu_lock protects below: * - per-IOMMU IOTLB caches
Signature · 1 line
2.7.4
RFC v10 14/25 intel_iommu: process PASID cache invalidation
In reply to Liu Yi LThis patch adds PASID cache invalidation handling. When guest enabled PASID usages (e.g. SVA), guest software should issue a proper PASID cache invalidation when caching-mode is exposed. This patch only adds the draft handling of pasid cache invalidation. Detailed handling will be added in subsequent patches.
---
rfcv4 (v1) -> rfcv5 (v2):
*) remove vtd_pasid_cache_gsi(), vtd_pasid_cache_psi()
and vtd_pasid_cache_dsi()
--- hw/i386/intel_iommu.c | 40 +++++++++++++++++++++++++++++++++++----- hw/i386/intel_iommu_internal.h | 12 ++++++++++++ hw/i386/trace-events | 3 +++ 3 files changed, 50 insertions(+), 5 deletions(-)
hw/i386/intel_iommu.c +35 -5
@@ -2395,6 +2395,37 @@ static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc)return true; } +static bool vtd_process_pasid_desc (IntelIOMMUState *s, +VTDInvDesc *inv_desc) +{ +if ((inv_desc->val[0 ] &VTD_INV_DESC_PASIDC_RSVD_VAL0 ) || + (inv_desc->val[1 ] &VTD_INV_DESC_PASIDC_RSVD_VAL1 ) || + (inv_desc->val[2 ] &VTD_INV_DESC_PASIDC_RSVD_VAL2 ) || + (inv_desc->val[3 ] &VTD_INV_DESC_PASIDC_RSVD_VAL3 )) { +error_report_once ("non-zero-field-in-pc_inv_desc hi: 0x%" PRIx64 +" lo: 0x%" PRIx64, inv_desc->val[1 ], inv_desc->val[0 ]); +return false; + } + +switch (inv_desc->val[0 ] &VTD_INV_DESC_PASIDC_G ) { +case VTD_INV_DESC_PASIDC_DSI : +break ; + +case VTD_INV_DESC_PASIDC_PASID_SI : +break ; + +case VTD_INV_DESC_PASIDC_GLOBAL : +break ; + +default : +error_report_once ("invalid-inv-granu-in-pc_inv_desc hi: 0x%" PRIx64 +" lo: 0x%" PRIx64, inv_desc->val[1 ], inv_desc->val[0 ]); +return false; + } + +return true; +} +static bool vtd_process_inv_iec_desc (IntelIOMMUState *s,VTDInvDesc *inv_desc) {
@@ -2501,12 +2532,11 @@ static bool vtd_process_inv_desc(IntelIOMMUState *s) }break ; -/* -* TODO: the entity of below two cases will be implemented in future series. -* To make guest (which integrates scalable mode support patch set in -* iommu driver) work, just return true is enough so far. -*/ case VTD_INV_DESC_PC : +trace_vtd_inv_desc ("pasid-cache" , inv_desc.val[1 ], inv_desc.val[0 ]); +if (!vtd_process_pasid_desc (s, &inv_desc)) { +return false; + }break ;case VTD_INV_DESC_PIOTLB :
hw/i386/intel_iommu_internal.h +12 -0
@@ -445,6 +445,18 @@ typedef union VTDInvDesc VTDInvDesc; (0x3ffff800ULL | ~(VTD_HAW_MASK(aw) | VTD_SL_IGN_COM | VTD_SL_TM)) : \ (0x3ffff800ULL | ~(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_VAL0 0xfff000000000ffc0ULL +#define VTD_INV_DESC_PASIDC_RSVD_VAL1 0xffffffffffffffffULL +#define VTD_INV_DESC_PASIDC_RSVD_VAL2 0xffffffffffffffffULL +#define VTD_INV_DESC_PASIDC_RSVD_VAL3 0xffffffffffffffffULL + +#define VTD_INV_DESC_PASIDC_DSI (0ULL << 4) +#define VTD_INV_DESC_PASIDC_PASID_SI (1ULL << 4) +#define VTD_INV_DESC_PASIDC_GLOBAL (3ULL << 4) +/* Information about page-selective IOTLB invalidate */ struct VTDIOTLBPageInvInfo {uint16_t domain_id;
hw/i386/trace-events +3 -0
@@ -22,6 +22,9 @@ vtd_inv_qi_head(uint16_t head) "read head %d" vtd_inv_qi_tail(uint16_t head) "write tail %d" vtd_inv_qi_fetch(void) "" vtd_context_cache_reset(void) "" +vtd_pasid_cache_gsi(void) "" +vtd_pasid_cache_dsi(uint16_t domain) "Domian slective PC invalidation domain 0x%"PRIx16 +vtd_pasid_cache_psi(uint16_t domain, uint32_t pasid) "PASID slective PC invalidation domain 0x%"PRIx16" pasid 0x%"PRIx32 vtd_re_not_present(uint8_t bus) "Root entry bus %"PRIu8" not present" vtd_ce_not_present(uint8_t bus, uint8_t devfn) "Context entry bus %"PRIu8" devfn %"PRIu8" not present" vtd_iotlb_page_hit(uint16_t sid, uint64_t addr, uint64_t slpte, uint16_t domain) "IOTLB page hit sid 0x%"PRIx16" iova 0x%"PRIx64" slpte 0x%"PRIx64" domain 0x%"PRIx16
Signature · 1 line
2.7.4
RFC v10 15/25 intel_iommu: add PASID cache management infrastructure
In reply to Liu Yi LThis patch adds a PASID cache management infrastructure based on new added structure VTDPASIDAddressSpace, which is used to track the PASID usage and future PASID tagged DMA address translation support in vIOMMU.
struct VTDPASIDAddressSpace {
VTDBus *vtd_bus;
uint8_t devfn;
AddressSpace as;
uint32_t pasid;
IntelIOMMUState *iommu_state;
VTDContextCacheEntry context_cache_entry;
QLIST_ENTRY(VTDPASIDAddressSpace) next;
VTDPASIDCacheEntry pasid_cache_entry;
};Ideally, a VTDPASIDAddressSpace instance is created when a PASID is bound with a DMA AddressSpace. Intel VT-d spec requires guest software to issue pasid cache invalidation when bind or unbind a pasid with an address space under caching-mode. However, as VTDPASIDAddressSpace instances also act as pasid cache in this implementation, its creation also happens during vIOMMU PASID tagged DMA translation. The creation in this path will not be added in this patch since no PASID-capable emulated devices for now.
The implementation in this patch manages VTDPASIDAddressSpace instances per PASID+BDF (lookup and insert will use PASID and BDF) since Intel VT-d spec allows per-BDF PASID Table. When a guest bind a PASID with an AddressSpace, QEMU will capture the guest pasid selective pasid cache invalidation, and allocate remove a VTDPASIDAddressSpace instance per the invalidation reasons:
*) a present pasid entry moved to non-present *) a present pasid entry to be a present entry *) a non-present pasid entry moved to present
vIOMMU emulator could figure out the reason by fetching latest guest pasid entry.
---
rfcv4 (v1) -> rfcv5 (v2):
*) merged this patch with former replay binding patch, makes
PSI/DSI/GSI use the unified function to do cache invalidation and pasid binding replay.
*) dropped pasid_cache_gen in both iommu_state and vtd_pasid_as
as it is not necessary so far, we may want it when one day initroduce emulated SVA-capable device.
--- hw/i386/intel_iommu.c | 464 +++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 21 ++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 24 +++ 4 files changed, 510 insertions(+)
hw/i386/intel_iommu.c +464 -0
@@ -40,6 +40,7 @@#include "kvm_i386.h" #include "migration/vmstate.h" #include "trace.h" +#include "qemu/jhash.h" /* context entry operations */ #define VTD_CE_GET_RID2PASID (ce) \
@@ -65,6 +66,8 @@static void vtd_address_space_refresh_all (IntelIOMMUState *s);static void vtd_address_space_unmap (VTDAddressSpace *as,IOMMUNotifier *n); +static void vtd_pasid_cache_reset (IntelIOMMUState *s); +static void vtd_panic_require_caching_mode (void ) { error_report("We need to set caching-mode=on for intel-iommu to enable "
@@ -276,6 +279,7 @@ static void vtd_reset_caches(IntelIOMMUState *s)vtd_iommu_lock (s);vtd_reset_iotlb_locked (s);vtd_reset_context_cache_locked (s); +vtd_pasid_cache_reset (s);vtd_iommu_unlock (s); }
@@ -686,6 +690,16 @@ static inline bool vtd_pe_type_check(X86IOMMUState *x86_iommu,return true; } +static inline uint16_t vtd_pe_get_domain_id (VTDPASIDEntry *pe) +{ +return VTD_SM_PASID_ENTRY_DID ((pe)->val[1 ]); +} + +static inline uint32_t vtd_sm_ce_get_pdt_entry_num (VTDContextEntry *ce) +{ +return 1U << (VTD_SM_CONTEXT_ENTRY_PDTS (ce->val[0 ]) +7 ); +} +static inline bool vtd_pdire_present (VTDPASIDDirEntry *pdire) {return pdire->val &1 ;
@@ -2395,9 +2409,443 @@ static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc)return true; } +static inline void vtd_init_pasid_key (uint32_t pasid, +uint16_t sid, +struct pasid_key *key) +{ + key->pasid = pasid; + key->sid = sid; +} + +static guint vtd_pasid_as_key_hash (gconstpointer v) +{ +struct pasid_key *key = (struct pasid_key *)v; +uint32_t a, b, c; + +/* Jenkins hash */ + a = b = c =JHASH_INITVAL +sizeof (*key); + a += key->sid; + b +=extract32 (key->pasid,0 ,16 ); + c +=extract32 (key->pasid,16 ,16 ); + +__jhash_mix (a, b, c); +__jhash_final (a, b, c); + +return c; +} + +static gboolean vtd_pasid_as_key_equal (gconstpointer v1,gconstpointer v2) +{ +const struct pasid_key *k1 = v1; +const struct pasid_key *k2 = v2; + +return (k1->pasid == k2->pasid) && (k1->sid == k2->sid); +} + +static inline int vtd_dev_get_pe_from_pasid (IntelIOMMUState *s, +uint8_t bus_num, +uint8_t devfn, +uint32_t pasid, +VTDPASIDEntry *pe) +{ +VTDContextEntry ce; +int ret; +dma_addr_t pasid_dir_base; + +if (!s->root_scalable) { +return -VTD_FR_PASID_TABLE_INV ; + } + + ret =vtd_dev_to_context_entry (s, bus_num, devfn, &ce); +if (ret) { +return ret; + } + + pasid_dir_base =VTD_CE_GET_PASID_DIR_TABLE (&ce); + ret =vtd_get_pe_from_pasid_table (s, + pasid_dir_base, pasid, pe); + +return ret; +} + +static bool vtd_pasid_entry_compare (VTDPASIDEntry *p1,VTDPASIDEntry *p2) +{ +return !memcmp (p1, p2,sizeof (*p1)); +} + +/** +* This function fills in the pasid entry in &vtd_pasid_as. Caller +* of this function should hold iommu_lock. +*/ +static void vtd_fill_pe_in_cache (IntelIOMMUState *s, +VTDPASIDAddressSpace *vtd_pasid_as, +VTDPASIDEntry *pe) +{ +VTDPASIDCacheEntry *pc_entry = &vtd_pasid_as->pasid_cache_entry; + +if (vtd_pasid_entry_compare (pe, &pc_entry->pasid_entry)) { +/* No need to go further as cached pasid entry is latest */ +return ; + } + + pc_entry->pasid_entry = *pe; +/* +* TODO: +* - send pasid bind to host for passthru devices +*/ +} + +/** +* This function is used to clear cached pasid entry in vtd_pasid_as +* instances. Caller of this function should hold iommu_lock. +*/ +static gboolean vtd_flush_pasid (gpointer key,gpointer value, +gpointer user_data) +{ +VTDPASIDCacheInfo *pc_info = user_data; +VTDPASIDAddressSpace *vtd_pasid_as = value; +IntelIOMMUState *s = vtd_pasid_as->iommu_state; +VTDPASIDCacheEntry *pc_entry = &vtd_pasid_as->pasid_cache_entry; +VTDBus *vtd_bus = vtd_pasid_as->vtd_bus; +VTDPASIDEntry pe; +uint16_t did; +uint32_t pasid; +uint16_t devfn; +int ret; + + did =vtd_pe_get_domain_id (&pc_entry->pasid_entry); + pasid = vtd_pasid_as->pasid; + devfn = vtd_pasid_as->devfn; + +switch (pc_info->type) { +case VTD_PASID_CACHE_FORCE_RESET : + goto remove; +case VTD_PASID_CACHE_PASIDSI : +if (pc_info->pasid != pasid) { +return false; + } +/* Fall through */ +case VTD_PASID_CACHE_DOMSI : +if (pc_info->domain_id != did) { +return false; + } +/* Fall through */ +case VTD_PASID_CACHE_GLOBAL_INV : +break ; +default : +error_report ("invalid pc_info->type" ); +abort (); + } + +/* +* pasid cache invalidation may indicate a present pasid +* entry to present pasid entry modification. To cover such +* case, vIOMMU emulator needs to fetch latest guest pasid +* entry and check cached pasid entry, then update pasid +* cache and send pasid bind/unbind to host properly. +*/ + ret =vtd_dev_get_pe_from_pasid (s,pci_bus_num (vtd_bus->bus), + devfn, pasid, &pe); +if (ret) { +/* +* No valid pasid entry in guest memory. e.g. pasid entry +* was modified to be either all-zero or non-present. Either +* case means existing pasid cache should be removed. +*/ + goto remove; + } + +vtd_fill_pe_in_cache (s, vtd_pasid_as, &pe); +/* +* TODO: +* - when pasid-base-iotlb(piotlb) infrastructure is ready, +* should invalidate QEMU piotlb togehter with this change. +*/ +return false; +remove: +/* +* TODO: +* - send pasid bind to host for passthru devices +* - when pasid-base-iotlb(piotlb) infrastructure is ready, +* should invalidate QEMU piotlb togehter with this change. +*/ +return true; +} + +/** +* This function finds or adds a VTDPASIDAddressSpace for a device +* when it is bound to a pasid. Caller of this function should hold +* iommu_lock. +*/ +static VTDPASIDAddressSpace *vtd_add_find_pasid_as (IntelIOMMUState *s, +VTDBus *vtd_bus, +int devfn, +uint32_t pasid) +{ +struct pasid_key key; +struct pasid_key *new_key; +VTDPASIDAddressSpace *vtd_pasid_as; +uint16_t sid; + + sid =vtd_make_source_id (pci_bus_num (vtd_bus->bus), devfn); +vtd_init_pasid_key (pasid, sid, &key); + vtd_pasid_as =g_hash_table_lookup (s->vtd_pasid_as, &key); + +if (!vtd_pasid_as) { + new_key =g_malloc0 (sizeof (*new_key)); +vtd_init_pasid_key (pasid, sid, new_key); +/* +* Initiate the vtd_pasid_as structure. +* +* This structure here is used to track the guest pasid +* binding and also serves as pasid-cache mangement entry. +* +* TODO: in future, if wants to support the SVA-aware DMA +* emulation, the vtd_pasid_as should have include +* AddressSpace to support DMA emulation. +*/ + vtd_pasid_as =g_malloc0 (sizeof (VTDPASIDAddressSpace)); + vtd_pasid_as->iommu_state = s; + vtd_pasid_as->vtd_bus = vtd_bus; + vtd_pasid_as->devfn = devfn; + vtd_pasid_as->pasid = pasid; +g_hash_table_insert (s->vtd_pasid_as, new_key, vtd_pasid_as); + } +return vtd_pasid_as; +} + +/** +* Caller of this function should hold iommu_lock. +*/ +static void vtd_sm_pasid_table_walk_one (IntelIOMMUState *s, +dma_addr_t pt_base, +int start, +int end, +VTDPASIDCacheInfo *info) +{ +VTDPASIDEntry pe; +int pasid = start; +int pasid_next; +VTDPASIDAddressSpace *vtd_pasid_as; + +while (pasid < end) { + pasid_next = pasid +1 ; + +if (!vtd_get_pe_in_pasid_leaf_table (s, pasid, pt_base, &pe) + &&vtd_pe_present (&pe)) { + vtd_pasid_as =vtd_add_find_pasid_as (s, + info->vtd_bus, info->devfn, pasid); +if ((info->type ==VTD_PASID_CACHE_DOMSI || + info->type ==VTD_PASID_CACHE_PASIDSI ) && + !(info->domain_id ==vtd_pe_get_domain_id (&pe))) { +/* +* VTD_PASID_CACHE_DOMSI and VTD_PASID_CACHE_PASIDSI +* requires domain ID check. If domain Id check fail, +* go to next pasid. +*/ + pasid = pasid_next; +continue ; + } +vtd_fill_pe_in_cache (s, vtd_pasid_as, &pe); + } + pasid = pasid_next; + } +} + +/* +* Currently, VT-d scalable mode pasid table is a two level table, +* this function aims to loop a range of PASIDs in a given pasid +* table to identify the pasid config in guest. +* Caller of this function should hold iommu_lock. +*/ +static void vtd_sm_pasid_table_walk (IntelIOMMUState *s, +dma_addr_t pdt_base, +int start, +int end, +VTDPASIDCacheInfo *info) +{ +VTDPASIDDirEntry pdire; +int pasid = start; +int pasid_next; +dma_addr_t pt_base; + +while (pasid < end) { + pasid_next = ((end - pasid) >VTD_PASID_TBL_ENTRY_NUM ) ? + (pasid +VTD_PASID_TBL_ENTRY_NUM ) : end; +if (!vtd_get_pdire_from_pdir_table (pdt_base, pasid, &pdire) + &&vtd_pdire_present (&pdire)) { + pt_base = pdire.val &VTD_PASID_TABLE_BASE_ADDR_MASK ; +vtd_sm_pasid_table_walk_one (s, pt_base, pasid, pasid_next, info); + } + pasid = pasid_next; + } +} + +static void vtd_replay_pasid_bind_for_dev (IntelIOMMUState *s, +int start,int end, +VTDPASIDCacheInfo *info) +{ +VTDContextEntry ce; +int bus_n, devfn; + + bus_n =pci_bus_num (info->vtd_bus->bus); + devfn = info->devfn; + +if (!vtd_dev_to_context_entry (s, bus_n, devfn, &ce)) { +uint32_t max_pasid; + + max_pasid =vtd_sm_ce_get_pdt_entry_num (&ce) *VTD_PASID_TBL_ENTRY_NUM ; +if (end > max_pasid) { + end = max_pasid; + } +vtd_sm_pasid_table_walk (s, +VTD_CE_GET_PASID_DIR_TABLE (&ce), + start, + end, + info); + } +} + +/** +* This function replay the guest pasid bindings to hots by +* walking the guest PASID table. This ensures host will have +* latest guest pasid bindings. Caller should hold iommu_lock. +*/ +static void vtd_replay_guest_pasid_bindings (IntelIOMMUState *s, +VTDPASIDCacheInfo *pc_info) +{ +VTDHostIOMMUContext *vtd_dev_icx; +int start =0 , end =VTD_HPASID_MAX ; +VTDPASIDCacheInfo walk_info; + +switch (pc_info->type) { +case VTD_PASID_CACHE_PASIDSI : + start = pc_info->pasid; + end = pc_info->pasid +1 ; +/* +* PASID selective invalidation is within domain, +* thus fall through. +*/ +case VTD_PASID_CACHE_DOMSI : +case VTD_PASID_CACHE_GLOBAL_INV : +/* loop all assigned devices */ +break ; +case VTD_PASID_CACHE_FORCE_RESET : +/* For force reset, no need to go further replay */ +return ; +default : +error_report ("invalid pc_info->type for replay" ); +abort (); + } + +/* +* In this replay, only needs to care about the devices which +* are backed by host IOMMU. For such devices, their vtd_dev_icx +* instances are in the s->vtd_dev_icx_list. For devices which +* are not backed byhost IOMMU, it is not necessary to replay +* the bindings since their cache could be re-created in the future +* DMA address transaltion. +*/ + walk_info = *pc_info; + QLIST_FOREACH(vtd_dev_icx, &s->vtd_dev_icx_list, next) { +/* vtd_bus|devfn fields are not identical with pc_info */ + walk_info.vtd_bus = vtd_dev_icx->vtd_bus; + walk_info.devfn = vtd_dev_icx->devfn; +vtd_replay_pasid_bind_for_dev (s, start, end, &walk_info); + } +} + +/** +* This function syncs the pasid bindings between guest and host. +* It includes updating the pasid cache in vIOMMU and updating the +* pasid bindings per guest's latest pasid entry presence. +*/ +static void vtd_pasid_cache_sync (IntelIOMMUState *s, +VTDPASIDCacheInfo *pc_info) +{ +/* +* Regards to a pasid cache invalidation, e.g. a PSI. +* it could be either cases of below: +* a) a present pasid entry moved to non-present +* b) a present pasid entry to be a present entry +* c) a non-present pasid entry moved to present +* +* Different invalidation granularity may affect different device +* scope and pasid scope. But for each invalidation granularity, +* it needs to do two steps to sync host and guest pasid binding. +* +* Here is the handling of a PSI: +* 1) loop all the existing vtd_pasid_as instances to update them +* according to the latest guest pasid entry in pasid table. +* this will make sure affected existing vtd_pasid_as instances +* cached the latest pasid entries. Also, during the loop, the +* host should be notified if needed. e.g. pasid unbind or pasid +* update. Should be able to cover case a) and case b). +* +* 2) loop all devices to cover case c) +* - For devices which have HostIOMMUContext instances, +* we loop them and check if guest pasid entry exists. If yes, +* it is case c), we update the pasid cache and also notify +* host. +* - For devices which have no HostIOMMUContext, it is not +* necessary to create pasid cache at this phase since it +* could be created when vIOMMU does DMA address translation. +* This is not yet implemented since there is no emulated +* pasid-capable devices today. If we have such devices in +* future, the pasid cache shall be created there. +* Other granularity follow the same steps, just with different scope +* +*/ + +vtd_iommu_lock (s); +/* Step 1: loop all the exisitng vtd_pasid_as instances */ +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, pc_info); + +/* +* Step 2: loop all the exisitng vtd_dev_icx instances. +* Ideally, needs to loop all devices to find if there is any new +* PASID binding regards to the PASID cache invalidation request. +* But it is enough to loop the devices which are backed by host +* IOMMU. For devices backed by vIOMMU (a.k.a emulated devices), +* if new PASID happened on them, their vtd_pasid_as instance could +* be created during future vIOMMU DMA translation. +*/ +vtd_replay_guest_pasid_bindings (s, pc_info); +vtd_iommu_unlock (s); +} + +/** +* Caller of this function should hold iommu_lock +*/ +static void vtd_pasid_cache_reset (IntelIOMMUState *s) +{ +VTDPASIDCacheInfo pc_info; + +trace_vtd_pasid_cache_reset (); + + pc_info.type =VTD_PASID_CACHE_FORCE_RESET ; + +/* +* Reset pasid cache is a big hammer, so use +* g_hash_table_foreach_remove which will free +* the vtd_pasid_as instances. Also, as a big +* hammer, use VTD_PASID_CACHE_FORCE_RESET to +* ensure all the vtd_pasid_as instances are +* dropped, meanwhile the change will be pass +* to host if HostIOMMUContext is available. +*/ +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); +} +static bool vtd_process_pasid_desc (IntelIOMMUState *s,VTDInvDesc *inv_desc) { +uint16_t domain_id; +uint32_t pasid; +VTDPASIDCacheInfo pc_info; + if ((inv_desc->val[0] & VTD_INV_DESC_PASIDC_RSVD_VAL0) || (inv_desc->val[1] & VTD_INV_DESC_PASIDC_RSVD_VAL1) || (inv_desc->val[2] & VTD_INV_DESC_PASIDC_RSVD_VAL2) ||
@@ -2407,14 +2855,26 @@ static bool vtd_process_pasid_desc(IntelIOMMUState *s,return false; } + domain_id =VTD_INV_DESC_PASIDC_DID (inv_desc->val[0 ]); + pasid =VTD_INV_DESC_PASIDC_PASID (inv_desc->val[0 ]); +switch (inv_desc->val[0 ] &VTD_INV_DESC_PASIDC_G ) {case VTD_INV_DESC_PASIDC_DSI : +trace_vtd_pasid_cache_dsi (domain_id); + pc_info.type =VTD_PASID_CACHE_DOMSI ; + pc_info.domain_id = domain_id;break ;case VTD_INV_DESC_PASIDC_PASID_SI : +/* PASID selective implies a DID selective */ + pc_info.type =VTD_PASID_CACHE_PASIDSI ; + pc_info.domain_id = domain_id; + pc_info.pasid = pasid;break ;case VTD_INV_DESC_PASIDC_GLOBAL : +trace_vtd_pasid_cache_gsi (); + pc_info.type =VTD_PASID_CACHE_GLOBAL_INV ;break ;default :
@@ -2423,6 +2883,7 @@ static bool vtd_process_pasid_desc(IntelIOMMUState *s,return false; } +vtd_pasid_cache_sync (s, &pc_info);return true; }
@@ -4112,6 +4573,9 @@ static void vtd_realize(DeviceState *dev, Error **errp) g_free, g_free); s->vtd_as_by_busptr =g_hash_table_new_full (vtd_uint64_hash, vtd_uint64_equal, g_free, g_free); + s->vtd_pasid_as =g_hash_table_new_full (vtd_pasid_as_key_hash, + vtd_pasid_as_key_equal, + g_free, g_free);vtd_init (s);sysbus_mmio_map (SYS_BUS_DEVICE (s),0 ,Q35_HOST_BRIDGE_IOMMU_ADDR );pci_setup_iommu (bus, &vtd_iommu_ops, dev);
hw/i386/intel_iommu_internal.h +21 -0
@@ -308,6 +308,7 @@ typedef enum VTDFaultReason {VTD_FR_IR_SID_ERR =0x26 ,/* Invalid Source-ID */ VTD_FR_PASID_TABLE_INV =0x58 ,/*Invalid PASID table entry */ +VTD_FR_PASID_ENTRY_P =0x59 ,/* The Present(P) field of pasidt-entry is 0 */ /* This is not a normal fault reason. We use this to indicate some faults * that are not referenced by the VT-d specification.
@@ -512,10 +513,29 @@ typedef struct VTDRootEntry VTDRootEntry;#define VTD_CTX_ENTRY_LEGACY_SIZE 16#define VTD_CTX_ENTRY_SCALABLE_SIZE 32 +#define VTD_SM_CONTEXT_ENTRY_PDTS (val) (((val) >> 9) & 0x3)#define VTD_SM_CONTEXT_ENTRY_RID2PASID_MASK 0xfffff#define VTD_SM_CONTEXT_ENTRY_RSVD_VAL0 (aw) (0x1e0ULL | ~VTD_HAW_MASK(aw))#define VTD_SM_CONTEXT_ENTRY_RSVD_VAL1 0xffffffffffe00000ULL +typedef enum VTDPCInvType { +/* force reset all */ +VTD_PASID_CACHE_FORCE_RESET =0 , +/* pasid cache invalidation rely on guest PASID entry */ +VTD_PASID_CACHE_GLOBAL_INV , +VTD_PASID_CACHE_DOMSI , +VTD_PASID_CACHE_PASIDSI , +}VTDPCInvType ; + +struct VTDPASIDCacheInfo { +VTDPCInvType type; +uint16_t domain_id; +uint32_t pasid; +VTDBus *vtd_bus; +uint16_t devfn; +}; +typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ; +/* PASID Table Related Definitions */ #define VTD_PASID_DIR_BASE_ADDR_MASK (~0xfffULL)#define VTD_PASID_TABLE_BASE_ADDR_MASK (~0xfffULL)
@@ -527,6 +547,7 @@ typedef struct VTDRootEntry VTDRootEntry;#define VTD_PASID_TABLE_BITS_MASK (0x3fULL)#define VTD_PASID_TABLE_INDEX (pasid) ((pasid) & VTD_PASID_TABLE_BITS_MASK)#define VTD_PASID_ENTRY_FPD (1ULL << 1)/* Fault Processing Disable */ +#define VTD_PASID_TBL_ENTRY_NUM (1ULL << 6)/* PASID Granular Translation Type Mask */ #define VTD_PASID_ENTRY_P 1ULL
hw/i386/trace-events +1 -0
@@ -23,6 +23,7 @@ vtd_inv_qi_tail(uint16_t head) "write tail %d" vtd_inv_qi_fetch(void) "" vtd_context_cache_reset(void) "" vtd_pasid_cache_gsi(void) "" +vtd_pasid_cache_reset(void) "" vtd_pasid_cache_dsi(uint16_t domain) "Domian slective PC invalidation domain 0x%"PRIx16 vtd_pasid_cache_psi(uint16_t domain, uint32_t pasid) "PASID slective PC invalidation domain 0x%"PRIx16" pasid 0x%"PRIx32 vtd_re_not_present(uint8_t bus) "Root entry bus %"PRIu8" not present"
include/hw/i386/intel_iommu.h +24 -0
@@ -65,6 +65,8 @@ typedef union VTD_IR_MSIAddress VTD_IR_MSIAddress;typedef struct VTDPASIDDirEntry VTDPASIDDirEntry ;typedef struct VTDPASIDEntry VTDPASIDEntry ;typedef struct VTDHostIOMMUContext VTDHostIOMMUContext ; +typedef struct VTDPASIDCacheEntry VTDPASIDCacheEntry ; +typedef struct VTDPASIDAddressSpace VTDPASIDAddressSpace ;/* Context-Entry */ struct VTDContextEntry {
@@ -97,6 +99,26 @@ struct VTDPASIDEntry {uint64_t val[8 ]; }; +struct pasid_key { +uint32_t pasid; +uint16_t sid; +}; + +struct VTDPASIDCacheEntry { +struct VTDPASIDEntry pasid_entry; +}; + +struct VTDPASIDAddressSpace { +VTDBus *vtd_bus; +uint8_t devfn; +AddressSpace as; +uint32_t pasid; +IntelIOMMUState *iommu_state; +VTDContextCacheEntry context_cache_entry; +QLIST_ENTRY (VTDPASIDAddressSpace) next; +VTDPASIDCacheEntry pasid_cache_entry; +}; +struct VTDAddressSpace {PCIBus *bus;uint8_t devfn;
@@ -267,6 +289,7 @@ struct IntelIOMMUState {GHashTable *vtd_as_by_busptr;/* VTDBus objects indexed by PCIBus* reference */ VTDBus *vtd_as_by_bus_num[VTD_PCI_BUS_MAX ];/* VTDBus objects indexed by bus number */ +GHashTable *vtd_pasid_as;/* VTDPASIDAddressSpace instances */ /* list of registered notifiers */ QLIST_HEAD(, VTDAddressSpace) vtd_as_with_notifiers;
@@ -292,6 +315,7 @@ struct IntelIOMMUState {* - per-IOMMU IOTLB caches * - context entry cache in VTDAddressSpace * - HostIOMMUContext pointer cached in vIOMMU +* - PASID cache in VTDPASIDAddressSpace */ QemuMutex iommu_lock; };
Signature · 1 line
2.7.4
RFC v10 16/25 vfio: add bind stage-1 page table support
In reply to Liu Yi LThis patch adds bind_stage1_pgtbl() definition in HostIOMMUContextClass, also adds corresponding implementation in VFIO. This is to expose a way for vIOMMU to setup dual stage DMA translation for passthru devices on hardware.
--- hw/iommu/host_iommu_context.c | 57 +++++++++++++++++++++++++++++++++- hw/vfio/common.c | 58 ++++++++++++++++++++++++++++++++++- include/hw/iommu/host_iommu_context.h | 19 +++++++++++- 3 files changed, 131 insertions(+), 3 deletions(-)
hw/iommu/host_iommu_context.c +56 -1
@@ -69,23 +69,78 @@ int host_iommu_ctx_pasid_free(HostIOMMUContext *iommu_ctx, uint32_t pasid)return hicxc->pasid_free (iommu_ctx, pasid); } +int host_iommu_ctx_bind_stage1_pgtbl (HostIOMMUContext *iommu_ctx, +struct iommu_gpasid_bind_data *bind) +{ +HostIOMMUContextClass *hicxc; + +if (!iommu_ctx) { +return -EINVAL ; + } + + hicxc =HOST_IOMMU_CONTEXT_GET_CLASS (iommu_ctx); +if (!hicxc) { +return -EINVAL ; + } + +if (!(iommu_ctx->flags &HOST_IOMMU_NESTING ) || + !hicxc->bind_stage1_pgtbl) { +return -EINVAL ; + } + +return hicxc->bind_stage1_pgtbl (iommu_ctx, bind); +} + +int host_iommu_ctx_unbind_stage1_pgtbl (HostIOMMUContext *iommu_ctx, +struct iommu_gpasid_bind_data *unbind) +{ +HostIOMMUContextClass *hicxc; + +if (!iommu_ctx) { +return -EINVAL ; + } + + hicxc =HOST_IOMMU_CONTEXT_GET_CLASS (iommu_ctx); +if (!hicxc) { +return -EINVAL ; + } + +if (!(iommu_ctx->flags &HOST_IOMMU_NESTING ) || + !hicxc->unbind_stage1_pgtbl) { +return -EINVAL ; + } + +return hicxc->unbind_stage1_pgtbl (iommu_ctx, unbind); +} +void host_iommu_ctx_init (void *_iommu_ctx,size_t instance_size,const char *mrtypename, -uint64_t flags) +uint64_t flags, +struct iommu_nesting_info *info) {HostIOMMUContext *iommu_ctx;object_initialize (_iommu_ctx, instance_size, mrtypename); iommu_ctx =HOST_IOMMU_CONTEXT (_iommu_ctx); iommu_ctx->flags = flags; + iommu_ctx->info =g_malloc0 (info->argsz); +memcpy (iommu_ctx->info, info, info->argsz); iommu_ctx->initialized = true; } +static void host_iommu_ctx_finalize_fn (Object *obj) +{ +HostIOMMUContext *iommu_ctx =HOST_IOMMU_CONTEXT (obj); + +g_free (iommu_ctx->info); +} +static const TypeInfo host_iommu_context_info = { .parent =TYPE_OBJECT , .name =TYPE_HOST_IOMMU_CONTEXT , .class_size =sizeof (HostIOMMUContextClass), .instance_size =sizeof (HostIOMMUContext), + .instance_finalize = host_iommu_ctx_finalize_fn, .abstract = true, };
hw/vfio/common.c +57 -1
79 lines not on this page · raw message
include/hw/iommu/host_iommu_context.h +18 -1
@@ -54,6 +54,16 @@ typedef struct HostIOMMUContextClass {/* Reclaim pasid from HostIOMMUContext (a.k.a. host software) */ int (*pasid_free)(HostIOMMUContext *iommu_ctx,uint32_t pasid); +/* +* Bind stage-1 page table to a hostIOMMU w/ dual stage +* DMA translation capability. +* @bind specifies the bind configurations. +*/ +int (*bind_stage1_pgtbl)(HostIOMMUContext *iommu_ctx, +struct iommu_gpasid_bind_data *bind); +/* Undo a previous bind. @unbind specifies the unbind info. */ +int (*unbind_stage1_pgtbl)(HostIOMMUContext *iommu_ctx, +struct iommu_gpasid_bind_data *unbind); } HostIOMMUContextClass;/*
@@ -62,17 +72,24 @@ typedef struct HostIOMMUContextClass {struct HostIOMMUContext {Object parent_obj;#define HOST_IOMMU_PASID_REQUEST (1ULL << 0) +#define HOST_IOMMU_NESTING (1ULL << 1)uint64_t flags; +struct iommu_nesting_info *info;bool initialized; };int host_iommu_ctx_pasid_alloc (HostIOMMUContext *iommu_ctx,uint32_t min,uint32_t max,uint32_t *pasid);int host_iommu_ctx_pasid_free (HostIOMMUContext *iommu_ctx,uint32_t pasid); +int host_iommu_ctx_bind_stage1_pgtbl (HostIOMMUContext *iommu_ctx, +struct iommu_gpasid_bind_data *bind); +int host_iommu_ctx_unbind_stage1_pgtbl (HostIOMMUContext *iommu_ctx, +struct iommu_gpasid_bind_data *unbind);void host_iommu_ctx_init (void *_iommu_ctx,size_t instance_size,const char *mrtypename, -uint64_t flags); +uint64_t flags, +struct iommu_nesting_info *info);void host_iommu_ctx_destroy (HostIOMMUContext *iommu_ctx);#endif
Signature · 1 line
2.7.4
RFC v10 17/25 intel_iommu: sync IOMMU nesting cap info for assigned devices
In reply to Liu Yi LFor assigned devices, Intel vIOMMU which wants to build DMA protection based on physical IOMMU nesting paging should check the IOMMU nesting support in host side. The host will return IOMMU nesting cap info to user-space (e.g. VFIO returns IOMMU nesting cap info for nesting type IOMMU). vIOMMU needs to check:
a) IOMMU model
b) 1st-level page table supports
c) address width
d) pasid support
This patch syncs the IOMMU nesting cap info when PCIe device (VFIO case) sets HostIOMMUContext to vIOMMU. If the host IOMMU nesting support is not compatible, vIOMMU should return failure to PCIe device.
--- hw/i386/intel_iommu.c | 105 +++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 18 +++++++ include/hw/i386/intel_iommu.h | 4 ++ 3 files changed, 127 insertions(+)
hw/i386/intel_iommu.c +105 -0
@@ -4103,6 +4103,82 @@ static int vtd_dev_get_iommu_attr(PCIBus *bus, void *opaque, int32_t devfn,return ret; } + +static bool vtd_check_nesting_info (IntelIOMMUState *s, +struct iommu_nesting_info *info, +struct iommu_nesting_info_vtd *vtd) +{ +return !((s->aw_bits != info->addr_width) || + ((s->host_cap ^ vtd->cap_reg) &VTD_CAP_MASK & s->host_cap) || + ((s->host_ecap ^ vtd->ecap_reg) &VTD_ECAP_MASK & s->host_ecap) || + (VTD_GET_PSS (s->host_ecap) != (info->pasid_bits -1 ))); +} + +/* Caller should hold iommu lock. */ +static bool vtd_sync_nesting_info (IntelIOMMUState *s, +struct iommu_nesting_info *info) +{ +struct iommu_nesting_info_vtd *vtd; +uint64_t cap, ecap; + + vtd = (struct iommu_nesting_info_vtd *) &info->vendor.vtd; + +if (s->cap_finalized) { +return vtd_check_nesting_info (s, info, vtd); + } + +if (s->aw_bits > info->addr_width) { +error_report ("User aw-bits: %u > host address width: %u" , + s->aw_bits, info->addr_width); +return false; + } + + cap = s->host_cap & vtd->cap_reg &VTD_CAP_MASK ; + s->host_cap &= ~VTD_CAP_MASK ; + s->host_cap |= cap; + + ecap = s->host_ecap & vtd->ecap_reg &VTD_ECAP_MASK ; + s->host_ecap &= ~VTD_ECAP_MASK ; + s->host_ecap |= ecap; + +if ((VTD_ECAP_PASID & s->host_ecap) && info->pasid_bits && + (VTD_GET_PSS (s->host_ecap) > (info->pasid_bits -1 ))) { + s->host_ecap &= ~VTD_ECAP_PSS_MASK ; + s->host_ecap |=VTD_ECAP_PSS (info->pasid_bits -1 ); + } +return true; +} + +/* +* virtual VT-d which wants nested needs to check the host IOMMU +* nesting cap info behind the assigned devices. Thus that vIOMMU +* could bind guest page table to host. +*/ +static bool vtd_check_iommu_ctx (IntelIOMMUState *s, +HostIOMMUContext *iommu_ctx) +{ +struct iommu_nesting_info *info = iommu_ctx->info; +uint32_t minsz, size; + +if (IOMMU_PASID_FORMAT_INTEL_VTD != info->format) { +error_report ("Format is not compatible for nesting!!!" ); +return false; + } + + size =sizeof (struct iommu_nesting_info_vtd ); + minsz = endof(struct iommu_nesting_info, flags); +if (size > (info->argsz - minsz)) { +/* +* QEMU may have been using new linux-headers/iommu.h than +* kernel supports, hence fail it. +*/ +error_report ("IOMMU nesting cap is not compatible!!!" ); +return false; + } + +return vtd_sync_nesting_info (s, info); +} +static int vtd_dev_set_iommu_context (PCIBus *bus,void *opaque,int devfn,HostIOMMUContext *iommu_ctx)
@@ -4117,6 +4193,11 @@ static int vtd_dev_set_iommu_context(PCIBus *bus, void *opaque,vtd_iommu_lock (s); +if (!vtd_check_iommu_ctx (s, iommu_ctx)) { +vtd_iommu_unlock (s); +return -ENOENT ; + } + vtd_dev_icx = vtd_bus->dev_icx[devfn];assert (!vtd_dev_icx);
@@ -4372,6 +4453,14 @@ static void vtd_init(IntelIOMMUState *s) s->ecap |=VTD_ECAP_SMTS |VTD_ECAP_SRS |VTD_ECAP_SLTS ; } +if (!s->cap_finalized) { + s->host_cap = s->cap; + s->host_ecap = s->ecap; + }else { + s->cap = s->host_cap; + s->ecap = s->host_ecap; + } +vtd_reset_caches (s);/* Define registers with default values and bit semantics */
@@ -4505,6 +4594,12 @@ static bool vtd_decide_config(IntelIOMMUState *s, Error **errp)return true; } +static void vtd_refresh_capability_reg (IntelIOMMUState *s) +{ +vtd_set_quad (s,DMAR_CAP_REG , s->cap); +vtd_set_quad (s,DMAR_ECAP_REG , s->ecap); +} +static int vtd_machine_done_notify_one (Object *child,void *unused) {IntelIOMMUState *iommu =INTEL_IOMMU_DEVICE (x86_iommu_get_default ());
@@ -4518,6 +4613,15 @@ static int vtd_machine_done_notify_one(Object *child, void *unused)vtd_panic_require_caching_mode (); } +vtd_iommu_lock (iommu); + iommu->cap = iommu->host_cap & iommu->cap; + iommu->ecap = iommu->host_ecap & iommu->ecap; +if (!iommu->cap_finalized) { + iommu->cap_finalized = true; + } + +vtd_refresh_capability_reg (iommu); +vtd_iommu_unlock (iommu);return 0 ; }
@@ -4549,6 +4653,7 @@ static void vtd_realize(DeviceState *dev, Error **errp)QLIST_INIT (&s->vtd_as_with_notifiers);QLIST_INIT (&s->vtd_dev_icx_list);qemu_mutex_init (&s->iommu_lock); + s->cap_finalized = false;memset (s->vtd_as_by_bus_num,0 ,sizeof (s->vtd_as_by_bus_num));memory_region_init_io (&s->csrmem,OBJECT (s), &vtd_mem_ops, s,"intel_iommu" ,DMAR_REG_SIZE );
hw/i386/intel_iommu_internal.h +18 -0
@@ -199,6 +199,24 @@#define VTD_ECAP_SMTS (1ULL << 43)#define VTD_ECAP_SLTS (1ULL << 46) +/* 1st level related caps */ +#define VTD_CAP_FL1GP (1ULL << 56) +#define VTD_CAP_FL5LP (1ULL << 60) +#define VTD_ECAP_PRS (1ULL << 29) +#define VTD_ECAP_ERS (1ULL << 30) +#define VTD_ECAP_SRS (1ULL << 31) +#define VTD_ECAP_EAFS (1ULL << 34) +#define VTD_ECAP_PSS (val) (((val) & 0x1fULL) << 35) +#define VTD_ECAP_PASID (1ULL << 40) + +#define VTD_GET_PSS (val) (((val) >> 35) & 0x1f) +#define VTD_ECAP_PSS_MASK (0x1fULL << 35) + +#define VTD_CAP_MASK (VTD_CAP_FL1GP | VTD_CAP_FL5LP) +#define VTD_ECAP_MASK (VTD_ECAP_PRS | VTD_ECAP_ERS | \ + VTD_ECAP_SRS | VTD_ECAP_EAFS | \ + VTD_ECAP_PASID) +/* CAP_REG */ /* (offset >> 4) << 24 */ #define VTD_CAP_FRO (DMAR_FRCD_REG_OFFSET << 20)
include/hw/i386/intel_iommu.h +4 -0
@@ -284,6 +284,9 @@ struct IntelIOMMUState {uint64_t cap;/* The value of capability reg */ uint64_t ecap;/* The value of extended capability reg */ +uint64_t host_cap;/* The value of host capability reg */ +uint64_t host_ecap;/* The value of host ext-capability reg */ +uint32_t context_cache_gen;/* Should be in [1,MAX] */ GHashTable *iotlb;/* IOTLB */
@@ -310,6 +313,7 @@ struct IntelIOMMUState {uint64_t vccap;/* The value of vcmd capability reg */ uint64_t vcrsp;/* Current value of VCMD RSP REG */ +bool cap_finalized;/* Whether VTD capability finalized */ /* * iommu_lock protects below: * - per-IOMMU IOTLB caches
Signature · 1 line
2.7.4
RFC v10 18/25 intel_iommu: bind/unbind guest page table to host
In reply to Liu Yi LThis patch captures the guest PASID table entry modifications and propagates the changes to host to setup dual stage DMA translation. The guest page table is configured as 1st level page table (GVA->GPA) whose translation result would further go through host VT-d 2nd level page table(GPA->HPA) under nested translation mode. This is the key part of vSVA support, and also a key to support IOVA over 1st- level page table for Intel VT-d in virtualization environment.
--- hw/i386/intel_iommu.c | 101 +++++++++++++++++++++++++++++++++++++++-- hw/i386/intel_iommu_internal.h | 18 ++++++++ 2 files changed, 114 insertions(+), 5 deletions(-)
hw/i386/intel_iommu.c +96 -5
@@ -41,6 +41,7 @@#include "migration/vmstate.h" #include "trace.h" #include "qemu/jhash.h" +#include <linux/iommu.h> /* context entry operations */ #define VTD_CE_GET_RID2PASID (ce) \
@@ -700,6 +701,16 @@ static inline uint32_t vtd_sm_ce_get_pdt_entry_num(VTDContextEntry *ce)return 1U << (VTD_SM_CONTEXT_ENTRY_PDTS (ce->val[0 ]) +7 ); } +static inline uint32_t vtd_pe_get_fl_aw (VTDPASIDEntry *pe) +{ +return 48 + ((pe->val[2 ] >>2 ) &VTD_SM_PASID_ENTRY_FLPM ) *9 ; +} + +static inline dma_addr_t vtd_pe_get_flpt_base (VTDPASIDEntry *pe) +{ +return pe->val[2 ] &VTD_SM_PASID_ENTRY_FLPTPTR ; +} +static inline bool vtd_pdire_present (VTDPASIDDirEntry *pdire) {return pdire->val &1 ;
@@ -1861,6 +1872,85 @@ static void vtd_context_global_invalidate(IntelIOMMUState *s)vtd_iommu_replay_all (s); } +/** +* Caller should hold iommu_lock. +*/ +static int vtd_bind_guest_pasid (IntelIOMMUState *s,VTDBus *vtd_bus, +int devfn,int pasid,VTDPASIDEntry *pe, +VTDPASIDOp op) +{ +VTDHostIOMMUContext *vtd_dev_icx; +HostIOMMUContext *iommu_ctx; +int ret =-1 ; + + vtd_dev_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_dev_icx) { +/* means no need to go further, e.g. for emulated devices */ +return 0 ; + } + + iommu_ctx = vtd_dev_icx->iommu_ctx; +if (!iommu_ctx) { +return -EINVAL ; + } + +switch (op) { +case VTD_PASID_BIND : + { +struct iommu_gpasid_bind_data *g_bind_data; + + g_bind_data =g_malloc0 (sizeof (*g_bind_data)); + + g_bind_data->argsz =sizeof (*g_bind_data); + g_bind_data->version =IOMMU_GPASID_BIND_VERSION_1 ; + g_bind_data->format =IOMMU_PASID_FORMAT_INTEL_VTD ; + g_bind_data->gpgd =vtd_pe_get_flpt_base (pe); + g_bind_data->addr_width =vtd_pe_get_fl_aw (pe); + g_bind_data->hpasid = pasid; + g_bind_data->gpasid = pasid; + g_bind_data->flags |=IOMMU_SVA_GPASID_VAL ; + g_bind_data->vendor.vtd.flags = + (VTD_SM_PASID_ENTRY_SRE_BIT (pe->val[2 ]) ? +IOMMU_SVA_VTD_GPASID_SRE :0 ) + | (VTD_SM_PASID_ENTRY_EAFE_BIT (pe->val[2 ]) ? +IOMMU_SVA_VTD_GPASID_EAFE :0 ) + | (VTD_SM_PASID_ENTRY_PCD_BIT (pe->val[1 ]) ? +IOMMU_SVA_VTD_GPASID_PCD :0 ) + | (VTD_SM_PASID_ENTRY_PWT_BIT (pe->val[1 ]) ? +IOMMU_SVA_VTD_GPASID_PWT :0 ) + | (VTD_SM_PASID_ENTRY_EMTE_BIT (pe->val[1 ]) ? +IOMMU_SVA_VTD_GPASID_EMTE :0 ) + | (VTD_SM_PASID_ENTRY_CD_BIT (pe->val[1 ]) ? +IOMMU_SVA_VTD_GPASID_CD :0 ); + g_bind_data->vendor.vtd.pat =VTD_SM_PASID_ENTRY_PAT (pe->val[1 ]); + g_bind_data->vendor.vtd.emt =VTD_SM_PASID_ENTRY_EMT (pe->val[1 ]); + ret =host_iommu_ctx_bind_stage1_pgtbl (iommu_ctx, g_bind_data); +g_free (g_bind_data); +break ; + } +case VTD_PASID_UNBIND : + { +struct iommu_gpasid_bind_data *g_unbind_data; + + g_unbind_data =g_malloc0 (sizeof (*g_unbind_data)); + + g_unbind_data->argsz =sizeof (*g_unbind_data); + g_unbind_data->version =IOMMU_GPASID_BIND_VERSION_1 ; + g_unbind_data->format =IOMMU_PASID_FORMAT_INTEL_VTD ; + g_unbind_data->hpasid = pasid; + ret =host_iommu_ctx_unbind_stage1_pgtbl (iommu_ctx, g_unbind_data); +g_free (g_unbind_data); +break ; + } +default : +error_report_once ("Unknown VTDPASIDOp!!!\n" ); +break ; + } + + +return ret; +} +/* Do a context-cache device-selective invalidation. * @func_mask: FM field after shifting */
@@ -2489,10 +2579,10 @@ static void vtd_fill_pe_in_cache(IntelIOMMUState *s, } pc_entry->pasid_entry = *pe; -/* -* TODO: -* - send pasid bind to host for passthru devices -*/ +vtd_bind_guest_pasid (s, vtd_pasid_as->vtd_bus, + vtd_pasid_as->devfn, + vtd_pasid_as->pasid, + pe,VTD_PASID_BIND ); }/**
@@ -2565,10 +2655,11 @@ static gboolean vtd_flush_pasid(gpointer key, gpointer value, remove:/* * TODO: -* - send pasid bind to host for passthru devices * - when pasid-base-iotlb(piotlb) infrastructure is ready, * should invalidate QEMU piotlb togehter with this change. */ +vtd_bind_guest_pasid (s, vtd_bus, devfn, + pasid,NULL ,VTD_PASID_UNBIND );return true; }
hw/i386/intel_iommu_internal.h +18 -0
@@ -536,6 +536,13 @@ typedef struct VTDRootEntry VTDRootEntry;#define VTD_SM_CONTEXT_ENTRY_RSVD_VAL0 (aw) (0x1e0ULL | ~VTD_HAW_MASK(aw))#define VTD_SM_CONTEXT_ENTRY_RSVD_VAL1 0xffffffffffe00000ULL +enum VTDPASIDOp { +VTD_PASID_BIND , +VTD_PASID_UNBIND , +VTD_OP_NUM +}; +typedef enum VTDPASIDOp VTDPASIDOp ; + typedef enum VTDPCInvType { /* force reset all */ VTD_PASID_CACHE_FORCE_RESET = 0,
@@ -578,6 +585,17 @@ typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo;#define VTD_SM_PASID_ENTRY_AW 7ULL/* Adjusted guest-address-width */ #define VTD_SM_PASID_ENTRY_DID (val) ((val) & VTD_DOMAIN_ID_MASK) +#define VTD_SM_PASID_ENTRY_FLPM 3ULL +#define VTD_SM_PASID_ENTRY_FLPTPTR (~0xfffULL) +#define VTD_SM_PASID_ENTRY_SRE_BIT (val) (!!((val) & 1ULL)) +#define VTD_SM_PASID_ENTRY_EAFE_BIT (val) (!!(((val) >> 7) & 1ULL)) +#define VTD_SM_PASID_ENTRY_PCD_BIT (val) (!!(((val) >> 31) & 1ULL)) +#define VTD_SM_PASID_ENTRY_PWT_BIT (val) (!!(((val) >> 30) & 1ULL)) +#define VTD_SM_PASID_ENTRY_EMTE_BIT (val) (!!(((val) >> 26) & 1ULL)) +#define VTD_SM_PASID_ENTRY_CD_BIT (val) (!!(((val) >> 25) & 1ULL)) +#define VTD_SM_PASID_ENTRY_PAT (val) (((val) >> 32) & 0xFFFFFFFFULL) +#define VTD_SM_PASID_ENTRY_EMT (val) (((val) >> 27) & 0x7ULL) +/* Second Level Page Translation Pointer*/ #define VTD_SM_PASID_ENTRY_SLPTPTR (~0xfffULL)
Signature · 1 line
2.7.4
RFC v10 19/25 intel_iommu: replay pasid binds after context cache invalidation
In reply to Liu Yi LThis patch replays guest pasid bindings after context cache invalidation. This is a behavior to ensure safety. Actually, programmer should issue pasid cache invalidation with proper granularity after issuing a context cache invalidation.
--- hw/i386/intel_iommu.c | 50 ++++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 1 + hw/i386/trace-events | 1 + 3 files changed, 52 insertions(+)
hw/i386/intel_iommu.c +50 -0
@@ -68,6 +68,10 @@ static void vtd_address_space_refresh_all(IntelIOMMUState *s);static void vtd_address_space_unmap (VTDAddressSpace *as,IOMMUNotifier *n);static void vtd_pasid_cache_reset (IntelIOMMUState *s); +static void vtd_pasid_cache_sync (IntelIOMMUState *s, +VTDPASIDCacheInfo *pc_info); +static void vtd_pasid_cache_devsi (IntelIOMMUState *s, +VTDBus *vtd_bus,uint16_t devfn);static void vtd_panic_require_caching_mode (void ) {
@@ -1853,7 +1857,10 @@ static void vtd_iommu_replay_all(IntelIOMMUState *s)static void vtd_context_global_invalidate (IntelIOMMUState *s) { +VTDPASIDCacheInfo pc_info; +trace_vtd_inv_desc_cc_global (); +/* Protects context cache */ vtd_iommu_lock (s); s->context_cache_gen++;
@@ -1870,6 +1877,9 @@ static void vtd_context_global_invalidate(IntelIOMMUState *s)* VT-d emulation codes. */ vtd_iommu_replay_all (s); + + pc_info.type =VTD_PASID_CACHE_GLOBAL_INV ; +vtd_pasid_cache_sync (s, &pc_info); }/**
@@ -2008,6 +2018,21 @@ static void vtd_context_device_invalidate(IntelIOMMUState *s,* happened. */ vtd_sync_shadow_page_table (vtd_as); +/* +* Per spec, context flush should also followed with PASID +* cache and iotlb flush. Regards to a device selective +* context cache invalidation: +* if (emaulted_device) +* invalidate pasid cahce and pasid-based iotlb +* else if (assigned_device) +* check if the device has been bound to any pasid +* invoke pasid_unbind regards to each bound pasid +* Here, we have vtd_pasid_cache_devsi() to invalidate pasid +* caches, while for piotlb in QEMU, we don't have it yet, so +* no handling. For assigned device, host iommu driver would +* flush piotlb when a pasid unbind is pass down to it. +*/ +vtd_pasid_cache_devsi (s, vtd_bus, devfn_it); } } }
@@ -2622,6 +2647,12 @@ static gboolean vtd_flush_pasid(gpointer key, gpointer value,/* Fall through */ case VTD_PASID_CACHE_GLOBAL_INV :break ; +case VTD_PASID_CACHE_DEVSI : +if (pc_info->vtd_bus != vtd_bus || + pc_info->devfn != devfn) { +return false; + } +break ;default :error_report ("invalid pc_info->type" );abort ();
@@ -2821,6 +2852,11 @@ static void vtd_replay_guest_pasid_bindings(IntelIOMMUState *s,case VTD_PASID_CACHE_GLOBAL_INV :/* loop all assigned devices */ break ; +case VTD_PASID_CACHE_DEVSI : + walk_info.vtd_bus = pc_info->vtd_bus; + walk_info.devfn = pc_info->devfn; +vtd_replay_pasid_bind_for_dev (s, start, end, &walk_info); +return ;case VTD_PASID_CACHE_FORCE_RESET :/* For force reset, no need to go further replay */ return ;
@@ -2906,6 +2942,20 @@ static void vtd_pasid_cache_sync(IntelIOMMUState *s,vtd_iommu_unlock (s); } +static void vtd_pasid_cache_devsi (IntelIOMMUState *s, +VTDBus *vtd_bus,uint16_t devfn) +{ +VTDPASIDCacheInfo pc_info; + +trace_vtd_pasid_cache_devsi (devfn); + + pc_info.type =VTD_PASID_CACHE_DEVSI ; + pc_info.vtd_bus = vtd_bus; + pc_info.devfn = devfn; + +vtd_pasid_cache_sync (s, &pc_info); +} +/** * Caller of this function should hold iommu_lock */
hw/i386/intel_iommu_internal.h +1 -0
@@ -548,6 +548,7 @@ typedef enum VTDPCInvType {VTD_PASID_CACHE_FORCE_RESET =0 ,/* pasid cache invalidation rely on guest PASID entry */ VTD_PASID_CACHE_GLOBAL_INV , +VTD_PASID_CACHE_DEVSI ,VTD_PASID_CACHE_DOMSI ,VTD_PASID_CACHE_PASIDSI , } VTDPCInvType;
hw/i386/trace-events +1 -0
@@ -26,6 +26,7 @@ vtd_pasid_cache_gsi(void) "" vtd_pasid_cache_reset(void) "" vtd_pasid_cache_dsi(uint16_t domain) "Domian slective PC invalidation domain 0x%"PRIx16 vtd_pasid_cache_psi(uint16_t domain, uint32_t pasid) "PASID slective PC invalidation domain 0x%"PRIx16" pasid 0x%"PRIx32 +vtd_pasid_cache_devsi(uint16_t devfn) "Dev selective PC invalidation dev: 0x%"PRIx16 vtd_re_not_present(uint8_t bus) "Root entry bus %"PRIu8" not present" vtd_ce_not_present(uint8_t bus, uint8_t devfn) "Context entry bus %"PRIu8" devfn %"PRIu8" not present" vtd_iotlb_page_hit(uint16_t sid, uint64_t addr, uint64_t slpte, uint16_t domain) "IOTLB page hit sid 0x%"PRIx16" iova 0x%"PRIx64" slpte 0x%"PRIx64" domain 0x%"PRIx16
Signature · 1 line
2.7.4
RFC v10 20/25 intel_iommu: do not pass down pasid bind for PASID #0
In reply to Liu Yi LRID_PASID field was introduced in VT-d 3.0 spec, it is used for DMA requests w/o PASID in scalable mode VT-d. It is also known as IOVA. And in VT-d 3.1 spec, there is definition on it:
"Implementations not supporting RID_PASID capability (ECAP_REG.RPS is 0b), use a PASID value of 0 to perform address translation for requests without PASID."
This patch adds a check against the PASIDs which are going to be bound to device. For PASID #0, it is not necessary to pass down pasid bind request for it since PASID #0 is used as RID_PASID for DMA requests without pasid. Further reason is current Intel vIOMMU supports gIOVA by shadowing guest 2nd level page table. However, in future, if guest IOMMU driver uses 1st level page table to store IOVA mappings, then guest IOVA support will also be done via nested translation. When gIOVA is over FLPT, then vIOMMU should pass down the pasid bind request for PASID #0 to host, host needs to bind the guest IOVA page table to a proper PASID. e.g. PASID value in RID_PASID field for PF/VF if ECAP_REG.RPS is clear or default PASID for ADI (Assignable Device Interface in Scalable IOV solution).
IOVA over FLPT support on Intel VT-d:
https://lore.kernel.org/linux-iommu/[email protected]/
--- hw/i386/intel_iommu.c | 10 ++++++++++ 1 file changed, 10 insertions(+)
hw/i386/intel_iommu.c +10 -0
@@ -1893,6 +1893,16 @@ static int vtd_bind_guest_pasid(IntelIOMMUState *s, VTDBus *vtd_bus,HostIOMMUContext *iommu_ctx;int ret =-1 ; +if (pasid <VTD_HPASID_MIN ) { +/* +* If pasid < VTD_HPASID_MIN, this pasid is not allocated +* from host. No need to pass down the changes on it to host. +* TODO: when IOVA over FLPT is ready, this switch should be +* refined. +*/ +return 0 ; + } + vtd_dev_icx = vtd_bus->dev_icx[devfn];if (!vtd_dev_icx) {/* means no need to go further, e.g. for emulated devices */
Signature · 1 line
2.7.4
RFC v10 21/25 vfio: add support for flush iommu stage-1 cache
In reply to Liu Yi LThis patch adds flush_stage1_cache() definition in HostIOMUContextClass. And adds corresponding implementation in VFIO. This is to expose a way for vIOMMU to flush stage-1 cache in host side since guest owns stage-1 translation structures in dual stage DMA translation configuration.
--- hw/iommu/host_iommu_context.c | 19 +++++++++++++++++++ hw/vfio/common.c | 24 ++++++++++++++++++++++++ include/hw/iommu/host_iommu_context.h | 8 ++++++++ 3 files changed, 51 insertions(+)
hw/iommu/host_iommu_context.c +19 -0
@@ -113,6 +113,25 @@ int host_iommu_ctx_unbind_stage1_pgtbl(HostIOMMUContext *iommu_ctx,return hicxc->unbind_stage1_pgtbl (iommu_ctx, unbind); } +int host_iommu_ctx_flush_stage1_cache (HostIOMMUContext *iommu_ctx, +struct iommu_cache_invalidate_info *cache) +{ +HostIOMMUContextClass *hicxc; + + hicxc =HOST_IOMMU_CONTEXT_GET_CLASS (iommu_ctx); + +if (!hicxc) { +return -EINVAL ; + } + +if (!(iommu_ctx->flags &HOST_IOMMU_NESTING ) || + !hicxc->flush_stage1_cache) { +return -EINVAL ; + } + +return hicxc->flush_stage1_cache (iommu_ctx, cache); +} + void host_iommu_ctx_init(void *_iommu_ctx, size_t instance_size, const char *mrtypename, uint64_t flags,
hw/vfio/common.c +24 -0
@@ -1275,6 +1275,29 @@ static int vfio_host_iommu_ctx_unbind_stage1_pgtbl(HostIOMMUContext *iommu_ctx,return ret; } +static int vfio_host_iommu_ctx_flush_stage1_cache (HostIOMMUContext *iommu_ctx, +struct iommu_cache_invalidate_info *cache) +{ +VFIOContainer *container =container_of (iommu_ctx, + VFIOContainer, iommu_ctx); +struct vfio_iommu_type1_nesting_op *op; +unsigned long argsz; +int ret =0 ; + + argsz =sizeof (*op) +sizeof (*cache); + op =g_malloc0 (argsz); + op->argsz = argsz; + op->flags =VFIO_IOMMU_NESTING_OP_CACHE_INVLD ; +memcpy (&op->data, cache,sizeof (*cache)); + +if (ioctl (container->fd,VFIO_IOMMU_NESTING_OP , op)) { + ret = -errno; +error_report ("%s: iommu cache flush failed: %m" , __func__); + } +g_free (op); +return ret; +} +/** * Get iommu info from host. Caller of this funcion should free * the memory pointed by the returned pointer stored in @info
@@ -2023,6 +2046,7 @@ static void vfio_host_iommu_context_class_init(ObjectClass *klass, hicxc->pasid_free = vfio_host_iommu_ctx_pasid_free; hicxc->bind_stage1_pgtbl = vfio_host_iommu_ctx_bind_stage1_pgtbl; hicxc->unbind_stage1_pgtbl = vfio_host_iommu_ctx_unbind_stage1_pgtbl; + hicxc->flush_stage1_cache = vfio_host_iommu_ctx_flush_stage1_cache; }static const TypeInfo vfio_host_iommu_context_info = {
include/hw/iommu/host_iommu_context.h +8 -0
@@ -64,6 +64,12 @@ typedef struct HostIOMMUContextClass {/* Undo a previous bind. @unbind specifies the unbind info. */ int (*unbind_stage1_pgtbl)(HostIOMMUContext *iommu_ctx,struct iommu_gpasid_bind_data *unbind); +/* +* Propagate stage-1 cache flush to host IOMMU, cache +* info specifid in @cache +*/ +int (*flush_stage1_cache)(HostIOMMUContext *iommu_ctx, +struct iommu_cache_invalidate_info *cache); } HostIOMMUContextClass;/*
@@ -85,6 +91,8 @@ int host_iommu_ctx_bind_stage1_pgtbl(HostIOMMUContext *iommu_ctx, struct iommu_gpasid_bind_data *bind);int host_iommu_ctx_unbind_stage1_pgtbl (HostIOMMUContext *iommu_ctx,struct iommu_gpasid_bind_data *unbind); +int host_iommu_ctx_flush_stage1_cache (HostIOMMUContext *iommu_ctx, +struct iommu_cache_invalidate_info *cache); void host_iommu_ctx_init(void *_iommu_ctx, size_t instance_size, const char *mrtypename,
Signature · 1 line
2.7.4
RFC v10 22/25 intel_iommu: process PASID-based iotlb invalidation
In reply to Liu Yi LThis patch adds the basic PASID-based iotlb (piotlb) invalidation support. piotlb is used during walking Intel VT-d 1st level page table. This patch only adds the basic processing. Detailed handling will be added in next patch.
--- hw/i386/intel_iommu.c | 53 ++++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 13 +++++++++++ 2 files changed, 66 insertions(+)
hw/i386/intel_iommu.c +53 -0
@@ -3038,6 +3038,55 @@ static bool vtd_process_pasid_desc(IntelIOMMUState *s,return true; } +static void vtd_piotlb_pasid_invalidate (IntelIOMMUState *s, +uint16_t domain_id, +uint32_t pasid) +{ +} + +static void vtd_piotlb_page_invalidate (IntelIOMMUState *s,uint16_t domain_id, +uint32_t pasid,hwaddr addr,uint8_t am, +bool ih) +{ +} + +static bool vtd_process_piotlb_desc (IntelIOMMUState *s, +VTDInvDesc *inv_desc) +{ +uint16_t domain_id; +uint32_t pasid; +uint8_t am; +hwaddr addr; + +if ((inv_desc->val[0 ] &VTD_INV_DESC_PIOTLB_RSVD_VAL0 ) || + (inv_desc->val[1 ] &VTD_INV_DESC_PIOTLB_RSVD_VAL1 )) { +error_report_once ("non-zero-field-in-piotlb_inv_desc hi: 0x%" PRIx64 +" lo: 0x%" PRIx64, inv_desc->val[1 ], inv_desc->val[0 ]); +return false; + } + + domain_id =VTD_INV_DESC_PIOTLB_DID (inv_desc->val[0 ]); + pasid =VTD_INV_DESC_PIOTLB_PASID (inv_desc->val[0 ]); +switch (inv_desc->val[0 ] &VTD_INV_DESC_IOTLB_G ) { +case VTD_INV_DESC_PIOTLB_ALL_IN_PASID : +vtd_piotlb_pasid_invalidate (s, domain_id, pasid); +break ; + +case VTD_INV_DESC_PIOTLB_PSI_IN_PASID : + am =VTD_INV_DESC_PIOTLB_AM (inv_desc->val[1 ]); + addr = (hwaddr )VTD_INV_DESC_PIOTLB_ADDR (inv_desc->val[1 ]); +vtd_piotlb_page_invalidate (s, domain_id, pasid, addr, am, +VTD_INV_DESC_PIOTLB_IH (inv_desc->val[1 ])); +break ; + +default : +error_report_once ("Invalid granularity in P-IOTLB desc hi: 0x%" PRIx64 +" lo: 0x%" PRIx64, inv_desc->val[1 ], inv_desc->val[0 ]); +return false; + } +return true; +} +static bool vtd_process_inv_iec_desc (IntelIOMMUState *s,VTDInvDesc *inv_desc) {
@@ -3152,6 +3201,10 @@ static bool vtd_process_inv_desc(IntelIOMMUState *s)break ;case VTD_INV_DESC_PIOTLB : +trace_vtd_inv_desc ("p-iotlb" , inv_desc.val[1 ], inv_desc.val[0 ]); +if (!vtd_process_piotlb_desc (s, &inv_desc)) { +return false; + }break ;case VTD_INV_DESC_WAIT :
hw/i386/intel_iommu_internal.h +13 -0
@@ -476,6 +476,19 @@ typedef union VTDInvDesc VTDInvDesc;#define VTD_INV_DESC_PASIDC_PASID_SI (1ULL << 4)#define VTD_INV_DESC_PASIDC_GLOBAL (3ULL << 4) +#define VTD_INV_DESC_PIOTLB_ALL_IN_PASID (2ULL << 4) +#define VTD_INV_DESC_PIOTLB_PSI_IN_PASID (3ULL << 4) + +#define VTD_INV_DESC_PIOTLB_RSVD_VAL0 0xfff000000000ffc0ULL +#define VTD_INV_DESC_PIOTLB_RSVD_VAL1 0xf80ULL + +#define VTD_INV_DESC_PIOTLB_PASID (val) (((val) >> 32) & 0xfffffULL) +#define VTD_INV_DESC_PIOTLB_DID (val) (((val) >> 16) & \ + VTD_DOMAIN_ID_MASK) +#define VTD_INV_DESC_PIOTLB_ADDR (val) ((val) & ~0xfffULL) +#define VTD_INV_DESC_PIOTLB_AM (val) ((val) & 0x3fULL) +#define VTD_INV_DESC_PIOTLB_IH (val) (((val) >> 6) & 0x1) +/* Information about page-selective IOTLB invalidate */ struct VTDIOTLBPageInvInfo {uint16_t domain_id;
Signature · 1 line
2.7.4
RFC v10 23/25 intel_iommu: propagate PASID-based iotlb invalidation to host
In reply to Liu Yi LThis patch propagates PASID-based iotlb invalidation to host.
Intel VT-d 3.0 supports nested translation in PASID granular. Guest SVA support could be implemented by configuring nested translation on specific PASID. This is also known as dual stage DMA translation.
Under such configuration, guest owns the GVA->GPA translation which is configured as first level page table in host side for a specific pasid, and host owns GPA->HPA translation. As guest owns first level translation table, piotlb invalidation should be propagated to host since host IOMMU will cache first level page table related mappings during DMA address translation.
This patch traps the guest PASID-based iotlb flush and propagate it to host.
---
rfcv4 (v1) -> rfcv5 (v2):
*) removed the valid check to vtd_pasid_as instance as rfcv5 ensures
all vtd_pasid_as instances in hash table should be valid.
--- hw/i386/intel_iommu.c | 113 +++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 7 +++ 2 files changed, 120 insertions(+)
hw/i386/intel_iommu.c +113 -0
@@ -3038,16 +3038,129 @@ static bool vtd_process_pasid_desc(IntelIOMMUState *s,return true; } +/** +* Caller of this function should hold iommu_lock. +*/ +static void vtd_invalidate_piotlb (IntelIOMMUState *s, +VTDBus *vtd_bus, +int devfn, +struct iommu_cache_invalidate_info *cache) +{ +VTDHostIOMMUContext *vtd_dev_icx; +HostIOMMUContext *iommu_ctx; + + vtd_dev_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_dev_icx) { + goto out; + } + iommu_ctx = vtd_dev_icx->iommu_ctx; +if (!iommu_ctx) { + goto out; + } +if (host_iommu_ctx_flush_stage1_cache (iommu_ctx, cache)) { +error_report ("Cache flush failed" ); + } +out: +return ; +} + +/** +* This function is a loop function for the s->vtd_pasid_as +* list with VTDPIOTLBInvInfo as execution filter. It propagates +* the piotlb invalidation to host. Caller of this function +* should hold iommu_lock. +*/ +static void vtd_flush_pasid_iotlb (gpointer key,gpointer value, +gpointer user_data) +{ +VTDPIOTLBInvInfo *piotlb_info = user_data; +VTDPASIDAddressSpace *vtd_pasid_as = value; +VTDPASIDCacheEntry *pc_entry = &vtd_pasid_as->pasid_cache_entry; +uint16_t did; + + did =vtd_pe_get_domain_id (&pc_entry->pasid_entry); + +if ((piotlb_info->domain_id == did) && + (piotlb_info->pasid == vtd_pasid_as->pasid)) { +vtd_invalidate_piotlb (vtd_pasid_as->iommu_state, + vtd_pasid_as->vtd_bus, + vtd_pasid_as->devfn, + piotlb_info->cache_info); + } + +/* +* TODO: needs to add QEMU piotlb flush when QEMU piotlb +* infrastructure is ready. For now, it is enough for passthru +* devices. +*/ +} +static void vtd_piotlb_pasid_invalidate (IntelIOMMUState *s,uint16_t domain_id,uint32_t pasid) { +VTDPIOTLBInvInfo piotlb_info; +struct iommu_cache_invalidate_info *cache_info; + + cache_info =g_malloc0 (sizeof (*cache_info)); + + cache_info->argsz =sizeof (*cache_info); + cache_info->version =IOMMU_CACHE_INVALIDATE_INFO_VERSION_1 ; + cache_info->cache =IOMMU_CACHE_INV_TYPE_IOTLB ; + cache_info->granularity =IOMMU_INV_GRANU_PASID ; + cache_info->granu.pasid_info.pasid = pasid; + cache_info->granu.pasid_info.flags =IOMMU_INV_PASID_FLAGS_PASID ; + + piotlb_info.domain_id = domain_id; + piotlb_info.pasid = pasid; + piotlb_info.cache_info = cache_info; + +vtd_iommu_lock (s); +/* +* Here loops all the vtd_pasid_as instances in s->vtd_pasid_as +* to find out the affected devices since piotlb invalidation +* should check pasid cache per architecture point of view. +*/ +g_hash_table_foreach (s->vtd_pasid_as, + vtd_flush_pasid_iotlb, &piotlb_info); +vtd_iommu_unlock (s); +g_free (cache_info); }static void vtd_piotlb_page_invalidate (IntelIOMMUState *s,uint16_t domain_id,uint32_t pasid,hwaddr addr,uint8_t am,bool ih) { +VTDPIOTLBInvInfo piotlb_info; +struct iommu_cache_invalidate_info *cache_info; + + cache_info =g_malloc0 (sizeof (*cache_info)); + + cache_info->argsz =sizeof (*cache_info); + cache_info->version =IOMMU_CACHE_INVALIDATE_INFO_VERSION_1 ; + cache_info->cache =IOMMU_CACHE_INV_TYPE_IOTLB ; + cache_info->granularity =IOMMU_INV_GRANU_ADDR ; + cache_info->granu.addr_info.flags =IOMMU_INV_ADDR_FLAGS_PASID ; + cache_info->granu.addr_info.flags |= ih ?IOMMU_INV_ADDR_FLAGS_LEAF :0 ; + cache_info->granu.addr_info.pasid = pasid; + cache_info->granu.addr_info.addr = addr; + cache_info->granu.addr_info.granule_size =1 << (12 + am); + cache_info->granu.addr_info.nb_granules =1 ; + + piotlb_info.domain_id = domain_id; + piotlb_info.pasid = pasid; + piotlb_info.cache_info = cache_info; + +vtd_iommu_lock (s); +/* +* Here loops all the vtd_pasid_as instances in s->vtd_pasid_as +* to find out the affected devices since piotlb invalidation +* should check pasid cache per architecture point of view. +*/ +g_hash_table_foreach (s->vtd_pasid_as, + vtd_flush_pasid_iotlb, &piotlb_info); +vtd_iommu_unlock (s); +g_free (cache_info); } static bool vtd_process_piotlb_desc(IntelIOMMUState *s,
hw/i386/intel_iommu_internal.h +7 -0
@@ -575,6 +575,13 @@ struct VTDPASIDCacheInfo { };typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ; +struct VTDPIOTLBInvInfo { +uint16_t domain_id; +uint32_t pasid; +struct iommu_cache_invalidate_info *cache_info; +}; +typedef struct VTDPIOTLBInvInfo VTDPIOTLBInvInfo ; +/* PASID Table Related Definitions */ #define VTD_PASID_DIR_BASE_ADDR_MASK (~0xfffULL)#define VTD_PASID_TABLE_BASE_ADDR_MASK (~0xfffULL)
Signature · 1 line
2.7.4
RFC v10 24/25 intel_iommu: process PASID-based Device-TLB invalidation
In reply to Liu Yi LThis patch adds an empty handling for PASID-based Device-TLB invalidation. For now it is enough as it is not necessary to propagate it to host for passthru device and also there is no emulated device has device tlb.
--- hw/i386/intel_iommu.c | 18 ++++++++++++++++++ hw/i386/intel_iommu_internal.h | 1 + 2 files changed, 19 insertions(+)
hw/i386/intel_iommu.c +18 -0
@@ -3213,6 +3213,17 @@ static bool vtd_process_inv_iec_desc(IntelIOMMUState *s,return true; } +static bool vtd_process_device_piotlb_desc (IntelIOMMUState *s, +VTDInvDesc *inv_desc) +{ +/* +* no need to handle it for passthru device, for emulated +* devices with device tlb, it may be required, but for now, +* return is enough +*/ +return true; +} +static bool vtd_process_device_iotlb_desc (IntelIOMMUState *s,VTDInvDesc *inv_desc) {
@@ -3334,6 +3345,13 @@ static bool vtd_process_inv_desc(IntelIOMMUState *s) }break ; +case VTD_INV_DESC_DEV_PIOTLB : +trace_vtd_inv_desc ("device-piotlb" , inv_desc.hi, inv_desc.lo); +if (!vtd_process_device_piotlb_desc (s, &inv_desc)) { +return false; + } +break ; +case VTD_INV_DESC_DEVICE :trace_vtd_inv_desc ("device" , inv_desc.hi, inv_desc.lo);if (!vtd_process_device_iotlb_desc (s, &inv_desc)) {
hw/i386/intel_iommu_internal.h +1 -0
@@ -405,6 +405,7 @@ typedef union VTDInvDesc VTDInvDesc;#define VTD_INV_DESC_WAIT 0x5/* Invalidation Wait Descriptor */ #define VTD_INV_DESC_PIOTLB 0x6/* PASID-IOTLB Invalidate Desc */ #define VTD_INV_DESC_PC 0x7/* PASID-cache Invalidate Desc */ +#define VTD_INV_DESC_DEV_PIOTLB 0x8/* PASID-based-DIOTLB inv_desc*/ #define VTD_INV_DESC_NONE 0/* Not an Invalidate Descriptor */ /* Masks for Invalidation Wait Descriptor*/
Signature · 1 line
2.7.4
RFC v10 25/25 intel_iommu: modify x-scalable-mode to be string option
In reply to Liu Yi LIntel VT-d 3.0 introduces scalable mode, and it has a bunch of capabilities related to scalable mode translation, thus there are multiple combinations. While this vIOMMU implementation wants simplify it for user by providing typical combinations. User could config it by "x-scalable-mode" option. The usage is as below:
"-device intel-iommu,x-scalable-mode=["legacy"|"modern"|"off"]"
- "legacy": gives support for SL page table
- "modern": gives support for FL page table, pasid, virtual command
- "off": no scalable mode support
- if not configured, means no scalable mode support, if not proper configured, will throw error
Note: this patch is supposed to be merged when the whole vSVA patch series were merged.
---
rfcv5 (v2) -> rfcv6:
*) reports want_nested to VFIO;
*) assert iommu_set/unset_iommu_context() if vIOMMU is not scalable modern.
--- hw/i386/intel_iommu.c | 39 +++++++++++++++++++++++++++++++++++---- hw/i386/intel_iommu_internal.h | 3 +++ include/hw/i386/intel_iommu.h | 2 ++ 3 files changed, 40 insertions(+), 4 deletions(-)
hw/i386/intel_iommu.c +35 -4
@@ -4050,7 +4050,7 @@ static Property vtd_properties[] = {DEFINE_PROP_UINT8 ("aw-bits" , IntelIOMMUState, aw_bits,VTD_HOST_ADDRESS_WIDTH ),DEFINE_PROP_BOOL ("caching-mode" , IntelIOMMUState, caching_mode, FALSE), -DEFINE_PROP_BOOL ("x-scalable-mode" , IntelIOMMUState, scalable_mode, FALSE), +DEFINE_PROP_STRING ("x-scalable-mode" , IntelIOMMUState, scalable_mode_str),DEFINE_PROP_BOOL ("dma-drain" , IntelIOMMUState, dma_drain, true), DEFINE_PROP_END_OF_LIST(), };
@@ -4419,6 +4419,7 @@ VTDAddressSpace *vtd_find_add_as(IntelIOMMUState *s, PCIBus *bus, int devfn)static int vtd_dev_get_iommu_attr (PCIBus *bus,void *opaque,int32_t devfn,IOMMUAttr attr,void *data) { +IntelIOMMUState *s = opaque;int ret =0 ;assert (0 <= devfn && devfn <PCI_DEVFN_MAX );
@@ -4428,8 +4429,7 @@ static int vtd_dev_get_iommu_attr(PCIBus *bus, void *opaque, int32_t devfn, {bool *pdata = data; -/* return false until vSVA is ready */ - *pdata = false; + *pdata = s->scalable_modern ? true : false;break ; }default :
@@ -4523,6 +4523,8 @@ static int vtd_dev_set_iommu_context(PCIBus *bus, void *opaque,VTDHostIOMMUContext *vtd_dev_icx;assert (0 <= devfn && devfn <PCI_DEVFN_MAX ); +/* only modern scalable supports set_ioimmu_context */ +assert (s->scalable_modern); vtd_bus =vtd_find_add_bus (s, bus);
@@ -4557,6 +4559,8 @@ static void vtd_dev_unset_iommu_context(PCIBus *bus, void *opaque, int devfn)VTDHostIOMMUContext *vtd_dev_icx;assert (0 <= devfn && devfn <PCI_DEVFN_MAX ); +/* only modern scalable supports unset_ioimmu_context */ +assert (s->scalable_modern); vtd_bus =vtd_find_add_bus (s, bus);
@@ -4784,8 +4788,13 @@ static void vtd_init(IntelIOMMUState *s) }/* TODO: read cap/ecap from host to decide which cap to be exposed. */ -if (s->scalable_mode) { +if (s->scalable_mode && !s->scalable_modern) { s->ecap |=VTD_ECAP_SMTS |VTD_ECAP_SRS |VTD_ECAP_SLTS ; + }else if (s->scalable_mode && s->scalable_modern) { + s->ecap |=VTD_ECAP_SMTS |VTD_ECAP_SRS |VTD_ECAP_PASID | +VTD_ECAP_FLTS |VTD_ECAP_PSS (VTD_PASID_SS ) | +VTD_ECAP_VCS ; + s->vccap |=VTD_VCCAP_PAS ; }if (!s->cap_finalized) {
@@ -4926,6 +4935,28 @@ static bool vtd_decide_config(IntelIOMMUState *s, Error **errp)return false; } +if (s->scalable_mode_str && + (strcmp (s->scalable_mode_str,"off" ) && +strcmp (s->scalable_mode_str,"modern" ) && +strcmp (s->scalable_mode_str,"legacy" ))) { +error_setg (errp,"Invalid x-scalable-mode config," +"Please use \"modern\", \"legacy\" or \"off\"" ); +return false; + } + +if (s->scalable_mode_str && + !strcmp (s->scalable_mode_str,"legacy" )) { + s->scalable_mode = true; + s->scalable_modern = false; + }else if (s->scalable_mode_str && + !strcmp (s->scalable_mode_str,"modern" )) { + s->scalable_mode = true; + s->scalable_modern = true; + }else { + s->scalable_mode = false; + s->scalable_modern = false; + } +return true; }
hw/i386/intel_iommu_internal.h +3 -0
@@ -197,7 +197,9 @@#define VTD_ECAP_MHMV (15ULL << 20)#define VTD_ECAP_SRS (1ULL << 31)#define VTD_ECAP_SMTS (1ULL << 43) +#define VTD_ECAP_VCS (1ULL << 44)#define VTD_ECAP_SLTS (1ULL << 46) +#define VTD_ECAP_FLTS (1ULL << 47)/* 1st level related caps */ #define VTD_CAP_FL1GP (1ULL << 56)
@@ -209,6 +211,7 @@#define VTD_ECAP_PSS (val) (((val) & 0x1fULL) << 35)#define VTD_ECAP_PASID (1ULL << 40) +#define VTD_PASID_SS (19)#define VTD_GET_PSS (val) (((val) >> 35) & 0x1f)#define VTD_ECAP_PSS_MASK (0x1fULL << 35)
include/hw/i386/intel_iommu.h +2 -0
@@ -263,6 +263,8 @@ struct IntelIOMMUState {bool caching_mode;/* RO - is cap CM enabled? */ bool scalable_mode;/* RO - is Scalable Mode supported? */ +char *scalable_mode_str;/* RO - admin's Scalable Mode config */ +bool scalable_modern;/* RO - is modern SM supported? */ dma_addr_t root;/* Current root table pointer */ bool root_scalable;/* Type of root table (scalable or not) */
Signature · 1 line
2.7.4
Hi Yi,
Quoted 17 lines · Liu Yi L
The high-level architecture for SVA virtualization is as below, the key design of vSVA support is to utilize the dual-stage IOMMU translation ( also known as IOMMU nesting translation) capability in host IOMMU.
.-------------. .---------------------------. | vIOMMU | | Guest process CR3, FL only| | | '---------------------------'.----------------/
| PASID Entry |--- PASID cache flush -
'-------------' |
| | V
| | CR3 in GPA
'-------------'
Guest
------| Shadow |--------------------------|--------
v v v
Host
Quoted 9 lines
.-------------. .----------------------.
| pIOMMU | | Bind FL for GVA-GPA |
| | '----------------------'
.----------------/ |
| PASID Entry | V (Nested xlate)
'----------------\.------------------------------.
| ||SL for GPA-HPA, default domain|
| | '------------------------------'
'-------------'A silly question: With nested mode, do we need to setup the second level mappings (GPA->HPA) for devices before vm startup? If so, can you please point me to the piece of code that achieves it?
On Arm, we setup the stage-2 mapping via the new prereg_listener [1].
[1] https://patchwork.kernel.org/project/qemu-devel/patch/[email protected]/
Thanks,
Zenghui