intel_iommu: expose Shared Virtual Addressing to VMs
v3 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. For IOMMU and VFIO changes, they are in separate series (listed in the "Related series").
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 RFC patchset is aiming for the phase 1 and phase 2, and works together with the VT-d driver[1] changes and VFIO changes[2].
Related series:
[1] [PATCH V9 00/10] Nested Shared Virtual Address (SVA) VT-d support:
https://lkml.org/lkml/2020/1/29/37 [PATCH 0/3] IOMMU user API enhancement: https://lkml.org/lkml/2020/1/29/45
[2] [RFC v3 0/8] vfio: expose virtual Shared Virtual Addressing to VMs
https://lkml.org/lkml/2020/1/29/255
There are roughly four parts:
1. Modify pci_setup_iommu() to set PCIIOMMUOps instead of setup PCIIOMMUFunc
2. Introduce DualStageIOMMUObject as abstract of host IOMMU. It provides method for vIOMMU emulators to communicate with host IOMMU. e.g. propagate guest page table binding to host IOMMU to setup dual-stage DMA translation in host IOMMU and flush iommu iotlb.
3. Introduce IOMMUContext as abstract of vIOMMU. It provides operations for VFIO to communicate with vIOMMU emulators. e.g. let vIOMMU emulators be aware of host IOMMU's dual-stage translation capability by registering DualStageIOMMUObject instances to vIOMMU emulators.
4. Setup dual-stage IOMMU translation for Intel vIOMMU. Includes
- Check IOMMU uAPI version compatibility and hardware compatibility which is preparation for setting up dual-stage DMA translation in host IOMMU.
- Propagate guest PASID allocation and free request to host.
- Propagate guest page table binding to host to setup dual-stage IOMMU DMA translation in host IOMMU.
- Propagate guest IOMMU cache invalidation to host to ensure iotlb correctness.
The complete QEMU set can be found in below link:
https://github.com/luxis1999/qemu.git: sva_vtd_v9_rfcv3
Complete kernel can be found in:
https://github.com/luxis1999/linux-vsva: vsva-linux-5.5-rc3
Changelog:
- 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 and addressed the TODOs mentioned in RFC v1.
RFC v1: https://patchwork.kernel.org/cover/11033657/
Eric Auger (1):
scripts/update-linux-headers: Import iommu.h
Liu Yi L (23):
hw/pci: modify pci_setup_iommu() to set PCIIOMMUOps hw/iommu: introduce DualStageIOMMUObject
hw/pci: introduce pci_device_iommu_context()
intel_iommu: provide get_iommu_context() callback header file update VFIO/IOMMU vSVA APIs
vfio: pass IOMMUContext into vfio_get_group()
vfio: check VFIO_TYPE1_NESTING_IOMMU support
vfio: register DualStageIOMMUObject to vIOMMU
vfio: get stage-1 pasid formats from Kernel vfio/common: add pasid_alloc/free support
intel_iommu: modify x-scalable-mode to be string option
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: bind/unbind guest page table to host
intel_iommu: replay guest pasid bindings 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
Peter Xu (1):
hw/iommu: introduce IOMMUContext
hw/Makefile.objs | 1 + 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 | 1208 ++++++++++++++++++++++++++++++++++- hw/i386/intel_iommu_internal.h | 119 ++++ hw/i386/trace-events | 6 + hw/iommu/Makefile.objs | 2 + hw/iommu/dual_stage_iommu.c | 101 +++ hw/iommu/iommu_context.c | 54 ++ hw/pci-host/designware.c | 6 +- hw/pci-host/ppce500.c | 6 +- hw/pci-host/prep.c | 6 +- hw/pci-host/sabre.c | 6 +- hw/pci/pci.c | 39 +- 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 | 244 ++++++- hw/vfio/pci.c | 3 +- hw/vfio/platform.c | 2 +- include/hw/i386/intel_iommu.h | 59 +- include/hw/iommu/dual_stage_iommu.h | 103 +++ include/hw/iommu/iommu_context.h | 61 ++ include/hw/pci/pci.h | 13 +- include/hw/pci/pci_bus.h | 2 +- include/hw/vfio/vfio-common.h | 6 +- linux-headers/linux/iommu.h | 372 +++++++++++ linux-headers/linux/vfio.h | 148 +++++ scripts/update-linux-headers.sh | 2 +- 33 files changed, 2568 insertions(+), 49 deletions(-) create mode 100644 hw/iommu/Makefile.objs create mode 100644 hw/iommu/dual_stage_iommu.c create mode 100644 hw/iommu/iommu_context.c create mode 100644 include/hw/iommu/dual_stage_iommu.h create mode 100644 include/hw/iommu/iommu_context.h create mode 100644 linux-headers/linux/iommu.h
Signature · 1 line
2.7.4
RFC v3 1/25 hw/pci: modify pci_setup_iommu() to set PCIIOMMUOps
From: Liu Yi L <[email protected]>
This 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.
--- 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/ppce500.c | 6 +++++- hw/pci-host/prep.c | 6 +++++- hw/pci-host/sabre.c | 6 +++++- hw/pci/pci.c | 12 +++++++----- hw/ppc/ppc440_pcix.c | 6 +++++- hw/ppc/spapr_pci.c | 6 +++++- hw/s390x/s390-pci-bus.c | 8 ++++++-- include/hw/pci/pci.h | 8 ++++++-- include/hw/pci/pci_bus.h | 2 +- 15 files changed, 75 insertions(+), 21 deletions(-)
hw/alpha/typhoon.c +5 -1
@@ -741,6 +741,10 @@ 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;
@@ -901,7 +905,7 @@ PCIBus *typhoon_init(ram_addr_t ram_size, ISABus **isa_bus, "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
@@ -342,6 +342,10 @@ 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;
@@ -436,7 +440,7 @@ 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
@@ -389,6 +389,10 @@ 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)
@@ -508,7 +512,7 @@ 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
@@ -1451,6 +1451,10 @@ 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,
@@ -1577,7 +1581,7 @@ 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" , errp);msi_init (&s->pci.dev,0 ,1 , true, false, errp);amdvi_init (s);
hw/i386/intel_iommu.c +5 -1
@@ -3722,6 +3722,10 @@ 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);
@@ -3833,7 +3837,7 @@ 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_set_parent_bus (DEVICE (&s->root),BUS (pci->bus));qdev_init_nofail (DEVICE (&s->root));
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 PCIIOMMU raven_iommu_ops = { + .get_address_space = raven_pcihost_set_iommu, +}; +static void raven_change_gpio (void *opaque,int n,int level) {PREPPCIState *s = opaque;
@@ -303,7 +307,7 @@ 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
@@ -112,6 +112,10 @@ 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) {
@@ -402,7 +406,7 @@ 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_create_multifunction(phb->bus, PCI_DEVFN(1, 0), true,
hw/pci/pci.c +7 -5
@@ -2644,7 +2644,7 @@ 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) { +while (iommu_bus && !iommu_bus->iommu_ops && iommu_bus->parent_dev) {PCIBus *parent_bus =pci_get_bus (iommu_bus->parent_dev);/*
@@ -2683,15 +2683,17 @@ 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_adress_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 ;
@@ -1950,7 +1954,7 @@ 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 ind_old, ind_new;
@@ -748,7 +752,7 @@ 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), &local_err);
@@ -919,7 +923,7 @@ 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), errp);
include/hw/pci/pci.h +6 -2
@@ -484,10 +484,14 @@ 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 v3 2/25 hw/iommu: introduce DualStageIOMMUObject
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
Currently, 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
As above, dual stage DMA translation offers two stage address mappings, which could have better DMA address translation support for passthru devices. This is also what vIOMMU developers are doing so far. Efforts includes vSVA enabling from Yi Liu and SMMUv3 Nested Stage Setup from Eric Auger.
https://www.spinics.net/lists/kvm/msg198556.html
https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg02842.html
Both efforts are aiming to expose a vIOMMU with dual stage hardware backed. As so, QEMU needs to have an explicit object to stand for the dual stage capability from hardware. Such object offers abstract for the dual stage DMA translation related operations, like:
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.
This patch introduces DualStageIOMMUObject to stand for the hardware dual stage DMA translation capability. PASID allocation/free are the first operation included in it, in future, there will be more operations like bind_stage1_pgtbl and invalidate_stage1_cache and etc.
--- hw/Makefile.objs | 1 + hw/iommu/Makefile.objs | 1 + hw/iommu/dual_stage_iommu.c | 59 +++++++++++++++++++++++++++++++++++++ include/hw/iommu/dual_stage_iommu.h | 59 +++++++++++++++++++++++++++++++++++++ 4 files changed, 120 insertions(+) create mode 100644 hw/iommu/Makefile.objs create mode 100644 hw/iommu/dual_stage_iommu.c create mode 100644 include/hw/iommu/dual_stage_iommu.h
hw/Makefile.objs +1 -0
@@ -40,6 +40,7 @@ devices-dirs-$(CONFIG_MEM_DEVICE) += mem/ devices-dirs-$(CONFIG_NUBUS) += nubus/ devices-dirs-y += semihosting/ devices-dirs-y += smbios/ +devices-dirs-y += iommu/ endif common-obj-y += $(devices-dirs-y)
hw/iommu/Makefile.objs new file +1 -0
@@ -0,0 +1 @@ +obj-y += dual_stage_iommu.o
hw/iommu/dual_stage_iommu.c new file +59 -0
59 lines not on this page · raw message
include/hw/iommu/dual_stage_iommu.h new file +59 -0
59 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [RFC v3 02/25] hw/iommu: introduce DualStageIOMMUObject
Quoted 42 lines · Yi L
From: Liu Yi L <[email protected]>
Currently, 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->GPAtranslation), and host IOMMU driver owns stage-2 translation structures (coversGPA->HPAtranslation). VMM is responsible to bind stage-1 translation structures to host, thus hardware could achieveGVA->GPAand thenGPA->HPAtranslation. 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.pdfAs above, dual stage DMA translation offers two stage address mappings, which could have better DMA address translation support for passthru devices. This is also what vIOMMU developers are doing so far. Efforts includes vSVA enabling from Yi Liu and SMMUv3 Nested Stage Setup from Eric Auger.
https://www.spinics.net/lists/kvm/msg198556.html
https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg02842.htmlBoth efforts are aiming to expose a vIOMMU with dual stage hardware backed. As so, QEMU needs to have an explicit object to stand for the dual stage capability from hardware. Such object offers abstract for the dual stage DMA translation related operations, like:
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.This patch introduces DualStageIOMMUObject to stand for the hardware dual stage DMA translation capability. PASID allocation/free are the first operation included in it, in future, there will be more operations like
bind_stage1_pgtblandinvalidate_stage1_cacheand etc.
Quoted 7 lines
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Signed-off-by Liu Yi L <[email protected]>
Several overall queries about this:
1) Since it's explicitly handling PASIDs, this seems a lot more specific to SVM than the name suggests. I'd suggest a rename.
2) Why are you hand rolling structures of pointers, rather than making this a QOM class or interface and putting those things into methods?
3) It's not really clear to me if this is for the case where both stages of translation are visible to the guest, or only one of them.
Signature · 4 lines
David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
RE: [RFC v3 02/25] hw/iommu: introduce DualStageIOMMUObject
Hi David,
Quoted 56 lines
From: David Gibson [mailto:[email protected]]
Sent: Friday, January 31, 2020 11:59 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 02/25] hw/iommu: introduce DualStageIOMMUObjectQuoted 42 lines · Yi L
From: Liu Yi L <[email protected]>
Currently, 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->GPAtranslation), and host IOMMU driver owns stage-2 translation structures (coversGPA->HPAtranslation). VMM is responsible to bind stage-1 translation structures to host, thus hardware could achieveGVA->GPAand thenGPA->HPAtranslation. 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.pdfAs above, dual stage DMA translation offers two stage address mappings, which could have better DMA address translation support for passthru devices. This is also what vIOMMU developers are doing so far. Efforts includes vSVA enabling from Yi Liu and SMMUv3 Nested Stage Setup from Eric Auger.
https://www.spinics.net/lists/kvm/msg198556.html
https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg02842.htmlBoth efforts are aiming to expose a vIOMMU with dual stage hardware backed. As so, QEMU needs to have an explicit object to stand for the dual stage capability from hardware. Such object offers abstract for the dual stage DMA translation related operations, like:
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.This patch introduces DualStageIOMMUObject to stand for the hardware dual stage DMA translation capability. PASID allocation/free are the first operation included in it, in future, there will be more operations like
bind_stage1_pgtblandinvalidate_stage1_cacheand etc.Quoted 7 lines
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Signed-off-by Liu Yi L <[email protected]>
Quoted 4 lines
Several overall queries about this:
1) Since it's explicitly handling PASIDs, this seems a lot more specific to SVM than the name suggests. I'd suggest a rename.
It is not specific to SVM in future. We have efforts to move guest IOVA support based on host IOMMU's dual-stage DMA translation capability. Then, guest IOVA support will also re-use the methods provided by this abstract layer. e.g. the bind_guest_pgtbl() and flush_iommu_iotlb().
For the naming, how about HostIOMMUContext? This layer is to provide explicit methods for setting up dual-stage DMA translation in host.
Quoted 3 lines
2) Why are you hand rolling structures of pointers, rather than making this a QOM class or interface and putting those things into methods?
Maybe the name is not proper. Although I named it as DualStageIOMMUObject, it is actually a kind of abstract layer we discussed in previous email. I think this is similar with VFIO_MAP/UNMAP. The difference is that VFIO_MAP/ UNMAP programs mappings to host iommu domain. While the newly added explicit method is to link guest page table to host iommu domain. VFIO_MAP/UNMAP is exposed to vIOMMU emulators via MemoryRegion layer. right? Maybe adding a similar abstract layer is enough. Is adding QOM really necessary for this case?
Quoted 3 lines
3) It's not really clear to me if this is for the case where both stages of translation are visible to the guest, or only one of them.
For this case, vIOMMU will only expose a single stage translation to VM. e.g. Intel VT-d, vIOMMU exposes first-level translation to guest. Hardware IOMMUs with the dual-stage translation capability lets guest own stage-1 translation structures and host owns the stage-2 translation structures. VMM is responsible to bind guest's translation structures to host and enable dual-stage translation. e.g. on Intel VT-d, config translation type to be NESTED.
Take guest SVM as an example, guest iommu driver owns the gVA->gPA mappings, which is treated as stage-1 translation from host point of view. Host itself owns the gPA->hPPA translation and called stage-2 translation when dual-stage translation is configured.
For guest IOVA, it is similar with guest SVM. Guest iommu driver owns the gIOVA->gPA mappings, which is treated as stage-1 translation. Host owns the gPA->hPA translation.
Regards,
Yi Liu
Re: [RFC v3 02/25] hw/iommu: introduce DualStageIOMMUObject
Quoted 58 lines · Yi L
Hi David,
Quoted 56 lines
From: David Gibson [mailto:[email protected]]
Sent: Friday, January 31, 2020 11:59 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 02/25] hw/iommu: introduce DualStageIOMMUObjectQuoted 49 lines · Yi L
From: Liu Yi L <[email protected]>
Currently, 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->GPAtranslation), and host IOMMU driver owns stage-2 translation structures (coversGPA->HPAtranslation). VMM is responsible to bind stage-1 translation structures to host, thus hardware could achieveGVA->GPAand thenGPA->HPAtranslation. 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.pdfAs above, dual stage DMA translation offers two stage address mappings, which could have better DMA address translation support for passthru devices. This is also what vIOMMU developers are doing so far. Efforts includes vSVA enabling from Yi Liu and SMMUv3 Nested Stage Setup from Eric Auger.
https://www.spinics.net/lists/kvm/msg198556.html
https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg02842.htmlBoth efforts are aiming to expose a vIOMMU with dual stage hardware backed. As so, QEMU needs to have an explicit object to stand for the dual stage capability from hardware. Such object offers abstract for the dual stage DMA translation related operations, like:
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.This patch introduces DualStageIOMMUObject to stand for the hardware dual stage DMA translation capability. PASID allocation/free are the first operation included in it, in future, there will be more operations like
bind_stage1_pgtblandinvalidate_stage1_cacheand etc.Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Signed-off-by Liu Yi L <[email protected]>
Quoted 8 lines
Quoted 4 lines
Several overall queries about this:
1) Since it's explicitly handling PASIDs, this seems a lot more specific to SVM than the name suggests. I'd suggest a rename.
It is not specific to SVM in future. We have efforts to move guest IOVA support based on host IOMMU's dual-stage DMA translation capability.
It's assuming the existence of pasids though, which is a rather more specific model than simply having two translation stages.
Quoted 6 lines
Then, guest IOVA support will also re-use the methods provided by this abstract layer. e.g. the
bind_guest_pgtbl()andflush_iommu_iotlb().For the naming, how about HostIOMMUContext? This layer is to provide explicit methods for setting up dual-stage DMA translation in host.
Uh.. maybe? I'm still having trouble figuring out what this object really represents.
Quoted 11 lines
Quoted 2 lines
2) Why are you hand rolling structures of pointers, rather than making this a QOM class or interface and putting those things into methods?
Maybe the name is not proper. Although I named it as DualStageIOMMUObject, it is actually a kind of abstract layer we discussed in previous email. I think this is similar with VFIO_MAP/UNMAP. The difference is that VFIO_MAP/ UNMAP programs mappings to host iommu domain. While the newly added explicit method is to link guest page table to host iommu domain. VFIO_MAP/UNMAP is exposed to vIOMMU emulators via MemoryRegion layer. right? Maybe adding a similar abstract layer is enough. Is adding QOM really necessary for this case?
Um... sorry, I'm having a lot of trouble making any sense of that.
Quoted 11 lines
Quoted 3 lines
3) It's not really clear to me if this is for the case where both stages of translation are visible to the guest, or only one of them.
For this case, vIOMMU will only expose a single stage translation to VM. e.g. Intel VT-d, vIOMMU exposes first-level translation to guest. Hardware IOMMUs with the dual-stage translation capability lets guest own stage-1 translation structures and host owns the stage-2 translation structures. VMM is responsible to bind guest's translation structures to host and enable dual-stage translation. e.g. on Intel VT-d, config translation type to be NESTED.
Ok, understood.
Quoted 8 lines
Take guest SVM as an example, guest iommu driver owns the
gVA->gPAmappings, which is treated as stage-1 translation from host point of view. Host itself owns thegPA->hPPAtranslation and called stage-2 translation when dual-stage translation is configured.For guest IOVA, it is similar with guest SVM. Guest iommu driver owns the
gIOVA->gPAmappings, which is treated as stage-1 translation. Host owns thegPA->hPAtranslation.
Ok, that makes sense. It's still not really clear to me which part of this setup this object represents.
Signature · 4 lines
David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
RFC v3 3/25 hw/iommu: introduce IOMMUContext
In reply to Liu, Yi LFrom: Peter Xu <[email protected]>
Currently, 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. Also there are efforts to make QEMU vIOMMU be backed by dual stage DMA address translation capability provided by hardware to have better address translation support for passthru devices.
As so, making vIOMMU be backed by dual stage translation capability requires QEMU vIOMMU to have a way to get aware of such hardware capability and also require a way to receive DMA address translation faults (e.g. I/O page request) from host as guest owns stage-1 translation structures in dual stage DAM address translation.
This patch adds IOMMUContext as an abstract of vIOMMU related operations. Like provide a way for passthru modules (e.g. VFIO) to register DualStageIOMMUObject instances. And in future, it is expected to offer support for receiving host DMA translation faults happened on stage-1 translation.
For more backgrounds, may refer to the discussion below, while there is also difference between the current implementation and original proposal. This patch introduces the IOMMUContext as an abstract layer for passthru module (e.g. VFIO) calls into vIOMMU. The first introduced interface is to make QEMU vIOMMU be aware of dual stage translation capability.
https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg05022.html
--- hw/iommu/Makefile.objs | 1 + hw/iommu/iommu_context.c | 54 +++++++++++++++++++++++++++++++++++ include/hw/iommu/iommu_context.h | 61 ++++++++++++++++++++++++++++++++++++++++ 3 files changed, 116 insertions(+) create mode 100644 hw/iommu/iommu_context.c create mode 100644 include/hw/iommu/iommu_context.h
hw/iommu/Makefile.objs +1 -0
@@ -1 +1,2 @@ obj-y += dual_stage_iommu.o +obj-y += iommu_context.o
hw/iommu/iommu_context.c new file +54 -0
54 lines not on this page · raw message
include/hw/iommu/iommu_context.h new file +61 -0
61 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
Quoted 21 lines · Yi L
From: Peter Xu <[email protected]>
Currently, 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. Also there are efforts to make QEMU vIOMMU be backed by dual stage DMA address translation capability provided by hardware to have better address translation support for passthru devices.
As so, making vIOMMU be backed by dual stage translation capability requires QEMU vIOMMU to have a way to get aware of such hardware capability and also require a way to receive DMA address translation faults (e.g. I/O page request) from host as guest owns stage-1 translation structures in dual stage DAM address translation.
This patch adds IOMMUContext as an abstract of vIOMMU related operations. Like provide a way for passthru modules (e.g. VFIO) to register DualStageIOMMUObject instances. And in future, it is expected to offer support for receiving host DMA translation faults happened on stage-1 translation.
Quoted 8 lines
For more backgrounds, may refer to the discussion below, while there is also difference between the current implementation and original proposal. This patch introduces the IOMMUContext as an abstract layer for passthru module (e.g. VFIO) calls into vIOMMU. The first introduced interface is to make QEMU vIOMMU be aware of dual stage translation capability.
https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg05022.html
Again, is there a reason for not making this a QOM class or interface?
I'm not very clear on the relationship betwen an IOMMUContext and a DualStageIOMMUObject. Can there be many IOMMUContexts to a DualStageIOMMUOBject? The other way around? Or is it just zero-or-one DualStageIOMMUObjects to an IOMMUContext?
Quoted 48 lines
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Signed-off-by Peter Xu <[email protected]>Signed-off-by Liu Yi L <[email protected]>--- hw/iommu/Makefile.objs | 1 + hw/iommu/iommu_context.c | 54 +++++++++++++++++++++++++++++++++++ include/hw/iommu/iommu_context.h | 61 ++++++++++++++++++++++++++++++++++++++++ 3 files changed, 116 insertions(+) create mode 100644 hw/iommu/iommu_context.c create mode 100644 include/hw/iommu/iommu_context.hhw/iommu/Makefile.objs +1
-0@@ -1 +1,2 @@ obj-y += dual_stage_iommu.o +obj-y += iommu_context.ohw/iommu/iommu_context.c new file +19
-0@@ -0,0 +1,54 @@ +/* +* QEMU abstract of vIOMMU context +* +* Copyright (C) 2020 Red Hat Inc. +* +* Authors: Peter Xu <[email protected]>, +* Liu Yi L <[email protected]> +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License as published by +* the Free Software Foundation; either version 2 of the License, or +* (at your option) any later version. + +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. + +* You should have received a copy of the GNU General Public License along
hw/iommu/iommu_context.c line 29
Quoted 10 lines
hw/iommu/iommu_context.c new file +10
-0+* with this program; if not, see <http://www.gnu.org/licenses/>. +*/ + +#include "qemu/osdep.h" +#include "hw/iommu/iommu_context.h" + +int iommu_context_register_ds_iommu (IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj) +{ +if (!iommu_ctx || !dsi_obj) {
Would this ever happen apart from a bug in the caller? If not it should be an assert().
Quoted 92 lines
hw/iommu/iommu_context.c new file +25
-0+return -ENOENT ; + } + +if (iommu_ctx->ops && iommu_ctx->ops->register_ds_iommu) { +return iommu_ctx->ops->register_ds_iommu (iommu_ctx, dsi_obj); + } +return -ENOENT ; +} + +void iommu_context_unregister_ds_iommu (IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj) +{ +if (!iommu_ctx || !dsi_obj) { +return ; + } + +if (iommu_ctx->ops && iommu_ctx->ops->unregister_ds_iommu) { + iommu_ctx->ops->unregister_ds_iommu (iommu_ctx, dsi_obj); + } +} + +void iommu_context_init (IOMMUContext *iommu_ctx,IOMMUContextOps *ops) +{ + iommu_ctx->ops = ops; +}include/hw/iommu/iommu_context.h new file +61
-0@@ -0,0 +1,61 @@ +/* +* QEMU abstraction of IOMMU Context +* +* Copyright (C) 2020 Red Hat Inc. +* +* Authors: Peter Xu <[email protected]>, +* Liu, Yi L <[email protected]> +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License as published by +* the Free Software Foundation; either version 2 of the License, or +* (at your option) any later version. + +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. + +* You should have received a copy of the GNU General Public License along +* with this program; if not, see <http://www.gnu.org/licenses/>. +*/ + +#ifndef HW_IOMMU_CONTEXT_H +#define HW_IOMMU_CONTEXT_H + +#include "qemu/queue.h" +#ifndef CONFIG_USER_ONLY +#include "exec/hwaddr.h" +#endif +#include "hw/iommu/dual_stage_iommu.h" + +typedef struct IOMMUContext IOMMUContext ; +typedef struct IOMMUContextOps IOMMUContextOps ; + +struct IOMMUContextOps { +/* +* Register DualStageIOMMUObject to vIOMMU thus vIOMMU +* is aware of dual stage translation capability, and +* also be able to setup dual stage translation via +* interfaces exposed by DualStageIOMMUObject. +*/ +int (*register_ds_iommu)(IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj); +void (*unregister_ds_iommu)(IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj); +}; + +/* +* This is an abstraction of IOMMU context. +*/ +struct IOMMUContext { +IOMMUContextOps *ops; +}; + +int iommu_context_register_ds_iommu (IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj); +void iommu_context_unregister_ds_iommu (IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj); +void iommu_context_init (IOMMUContext *iommu_ctx,IOMMUContextOps *ops); + +#endif
Signature · 4 lines
David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
RE: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
Hi David,
Quoted 34 lines
From: David Gibson [mailto:[email protected]]
Sent: Friday, January 31, 2020 12:07 PM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContextQuoted 28 lines · Yi L
From: Peter Xu <[email protected]>
Currently, 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. Also there are efforts to make QEMU vIOMMU be backed by dual stage DMA address translation capability provided by hardware to have better address translation support for passthru devices.
As so, making vIOMMU be backed by dual stage translation capability requires QEMU vIOMMU to have a way to get aware of such hardware capability and also require a way to receive DMA address translation faults (e.g. I/O page request) from host as guest owns stage-1 translation structures in dual stage DAM address translation.
This patch adds IOMMUContext as an abstract of vIOMMU related operations. Like provide a way for passthru modules (e.g. VFIO) to register DualStageIOMMUObject instances. And in future, it is expected to offer support for receiving host DMA translation faults happened on stage-1 translation.
For more backgrounds, may refer to the discussion below, while there is also difference between the current implementation and original proposal. This patch introduces the IOMMUContext as an abstract layer for passthru module (e.g. VFIO) calls into vIOMMU. The first introduced interface is to make QEMU vIOMMU be aware of dual stage translation capability.
Quoted 3 lines
Again, is there a reason for not making this a QOM class or interface?
I guess it is enough to make a simple abstract layer as explained in prior email. IOMMUContext is to provide explicit method for VFIO to call into vIOMMU emulators.
Quoted 5 lines
I'm not very clear on the relationship betwen an IOMMUContext and a DualStageIOMMUObject. Can there be many IOMMUContexts to a DualStageIOMMUOBject? The other way around? Or is it just zero-or-one DualStageIOMMUObjects to an IOMMUContext?
It is possible. As the below patch shows, DualStageIOMMUObject is per vfio container. IOMMUContext can be either per-device or shared across devices, it depends on vendor specific vIOMMU emulators.
[RFC v3 10/25] vfio: register DualStageIOMMUObject to vIOMMU
https://www.spinics.net/lists/kvm/msg205198.html
Take Intel vIOMMU as an example, there is a per device structure which includes IOMMUContext instance and a DualStageIOMMUObject pointer.
+struct VTDIOMMUContext {
+ VTDBus *vtd_bus; + uint8_t devfn; + IOMMUContext iommu_context; + DualStageIOMMUObject *dsi_obj; + IntelIOMMUState *iommu_state;
+};
https://www.spinics.net/lists/kvm/msg205196.html
I think this would leave space for vendor specific vIOMMU emulators to design their own relationship between an IOMMUContext and a DualStageIOMMUObject.
Quoted 3 lines
Quoted 3 lines
[...]
Quoted 4 lines
Quoted 4 lines
+ */
hw/iommu/iommu_context.c new file +3
-0+ +#include "qemu/osdep.h" +#include "hw/iommu/iommu_context.h"
Quoted 8 lines
hw/iommu/iommu_context.c line 29
Quoted 5 lines
hw/iommu/iommu_context.c new file +5
-0+ +int iommu_context_register_ds_iommu (IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj) +{ +if (!iommu_ctx || !dsi_obj) {Would this ever happen apart from a bug in the caller? If not it should be an
assert().
Got it, thanks, I'll check all other alike in this series and fix them in next version.
Thanks,
Yi Liu
Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
Quoted 8 lines · Yi L
Quoted 4 lines
I'm not very clear on the relationship betwen an IOMMUContext and a DualStageIOMMUObject. Can there be many IOMMUContexts to a DualStageIOMMUOBject? The other way around? Or is it just zero-or-one DualStageIOMMUObjects to an IOMMUContext?
It is possible. As the below patch shows, DualStageIOMMUObject is per vfio container. IOMMUContext can be either per-device or shared across devices, it depends on vendor specific vIOMMU emulators.
Is there an example when an IOMMUContext can be not per-device?
It makes sense to me to have an object that is per-container (in your case, the DualStageIOMMUObject, IIUC), then we can connect that object to a device. However I'm a bit confused on why we've got two abstract layers (the other one is IOMMUContext)? That was previously for the whole SVA new APIs, now it's all moved over to the other new object, then IOMMUContext only register/unregister... Can we put the reg/unreg procedures into DualStageIOMMUObject as well? Then we drop the IOMMUContext (or say, keep IOMMUContext and drop DualStageIOMMUObject but let IOMMUContext to be per-vfio-container, the major difference is the naming here, say, PASID allocation does not seem to be related to dual-stage at all).
Besides that, not sure I read it right... but even with your current series, the container->iommu_ctx will always only be bound to the first device created within that container, since you've got:
group = vfio_get_group(groupid, pci_device_iommu_address_space(pdev),
pci_device_iommu_context(pdev), errp);
And:
if (vfio_connect_container(group, as, iommu_ctx, errp)) {
error_prepend(errp, "failed to setup container for group %d: ",
groupid);
goto close_fd_exit;
}The iommu_ctx will be set to container->iommu_ctx if there's no existing container.
Quoted 18 lines
[RFC v3 10/25] vfio: register DualStageIOMMUObject to vIOMMU
https://www.spinics.net/lists/kvm/msg205198.htmlTake Intel vIOMMU as an example, there is a per device structure which includes IOMMUContext instance and a DualStageIOMMUObject pointer.
(quoted diff) +7
-0+struct VTDIOMMUContext { + VTDBus *vtd_bus; + uint8_t devfn; + IOMMUContext iommu_context; + DualStageIOMMUObject *dsi_obj; + IntelIOMMUState *iommu_state; +};https://www.spinics.net/lists/kvm/msg205196.html
I think this would leave space for vendor specific vIOMMU emulators to design their own relationship between an IOMMUContext and a DualStageIOMMUObject.
Signature · 1 line
Peter Xu
RE: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
Hi Peter,
Quoted 11 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 12:59 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContextQuoted 5 lines · Yi L
Quoted 4 lines
I'm not very clear on the relationship betwen an IOMMUContext and a DualStageIOMMUObject. Can there be many IOMMUContexts to a DualStageIOMMUOBject? The other way around? Or is it just zero-or-one DualStageIOMMUObjects to an IOMMUContext?
Quoted 5 lines
Quoted 3 lines
It is possible. As the below patch shows, DualStageIOMMUObject is per vfio container. IOMMUContext can be either per-device or shared across devices, it depends on vendor specific vIOMMU emulators.
Is there an example when an IOMMUContext can be not per-device?
No, I don’t have such example so far. But as IOMMUContext is got from pci_device_iommu_context(), in concept it possible to be not per-device. It is kind of leave to vIOMMU to decide if different devices could share a single IOMMUContext.
Quoted 6 lines
It makes sense to me to have an object that is per-container (in your case, the DualStageIOMMUObject, IIUC), then we can connect that object to a device. However I'm a bit confused on why we've got two abstract layers (the other one is IOMMUContext)? That was previously for the whole SVA new APIs, now it's all moved over to the other new object, then IOMMUContext only register/unregister...
Your understanding is correct. Actually, I also struggled on adding two abstract layer. But, you know, there are two function calling requirements around vSVA enabling. First one is explicit method for vIOMMU calls into VFIO (pasid allocation, bind guest page table, cache invalidate). Second one is explicit method for VFIO calls into vIOMMU (DMA fault/PRQ injection which is not included in this series yet, but will be upstreamed later). So I added the DualStageIOMMUObject to cover vIOMMU to VFIO callings, and IOMMUContext to cover VFIO to vIOMMU callings. As IOMMUContext covers VFIO to vIOMMU callings, so I made it include register/unregister.
Quoted 16 lines
Can we put the reg/unreg
procedures into DualStageIOMMUObject as well? Then we drop the IOMMUContext (or say, keep IOMMUContext and drop DualStageIOMMUObject but let IOMMUContext to be per-vfio-container, the major difference is the naming here, say, PASID allocation does not seem to be related to dual-stage at all).Besides that, not sure I read it right... but even with your current series, the
container->iommu_ctxwill always only be bound to the first device created within that container, since you've got:group = vfio_get_group(groupid,
pci_device_iommu_address_space(pdev),
pci_device_iommu_context(pdev), errp);And:
Quoted 8 lines
if (vfio_connect_container(group, as, iommu_ctx, errp)) { error_prepend(errp, "failed to setup container for group %d: ", groupid); goto close_fd_exit; }The
iommu_ctxwill be set tocontainer->iommu_ctxif there's no existing container.
yes, it's true. May need to add a iommu_ctx list in VFIO container or add check on the input iommu_ctx of vfio_get_group() if sticking on this direction.
While considering your suggestion on dropping one of the two abstract layers. I came up a new proposal as below:
We may drop the IOMMUContext in this series, and rename DualStageIOMMUObject to HostIOMMUContext, which is per-vfio-container. Add an interface in PCI layer(e.g. an callback in PCIDevice) to let vIOMMU get HostIOMMUContext. I think this could cover the requirement of providing explicit method for vIOMMU to call into VFIO and then program host IOMMU.
While for the requirement of VFIO to vIOMMU callings (e.g. PRQ), I think it could be done via PCI layer by adding an operation in PCIIOMMUOps. Thoughts?
Thanks,
Yi Liu
Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
On Wed, Feb 12, 2020 at 07:15:13AM +0000, Liu, Yi L wrote:
[...]
Quoted 11 lines
While considering your suggestion on dropping one of the two abstract layers. I came up a new proposal as below:
We may drop the IOMMUContext in this series, and rename DualStageIOMMUObject to HostIOMMUContext, which is per-vfio-container. Add an interface in PCI layer(e.g. an callback in PCIDevice) to let vIOMMU get HostIOMMUContext. I think this could cover the requirement of providing explicit method for vIOMMU to call into VFIO and then program host IOMMU.
While for the requirement of VFIO to vIOMMU callings (e.g. PRQ), I think it could be done via PCI layer by adding an operation in PCIIOMMUOps. Thoughts?
Hmm sounds good. :)
The thing is for the calls to the other direction (e.g. VFIO injecting faults to vIOMMU), that's neither per-container nor per-device, but per-vIOMMU. PCIIOMMUOps suites for that job I'd say, which is per-vIOMMU.
Let's see how it goes.
Signature · 1 line
Peter Xu
RE: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
Quoted 23 lines
From: Peter Xu <[email protected]>
Sent: Thursday, February 13, 2020 12:00 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContextOn Wed, Feb 12, 2020 at 07:15:13AM +0000, Liu, Yi L wrote:
[...]
Quoted 6 lines
While considering your suggestion on dropping one of the two abstract layers. I came up a new proposal as below:
We may drop the IOMMUContext in this series, and rename DualStageIOMMUObject to HostIOMMUContext, which is per-vfio-container. Add an interface in PCI layer(e.g. an callback in PCIDevice) to let vIOMMU get
HostIOMMUContext.
Quoted 5 lines
I think this could cover the requirement of providing explicit method for vIOMMU to call into VFIO and then program host IOMMU.
While for the requirement of VFIO to vIOMMU callings (e.g. PRQ), I think it could be done via PCI layer by adding an operation in PCIIOMMUOps.
Thoughts?
Quoted 7 lines
Hmm sounds good. :)
The thing is for the calls to the other direction (e.g. VFIO injecting faults to vIOMMU), that's neither per-container nor per-device, but per-vIOMMU. PCIIOMMUOps suites for that job I'd say, which is per-vIOMMU.
Let's see how it goes.
Thanks, let me get a new version by end-of this week.
Regards,
Yi Liu
Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
In reply to Liu, Yi LQuoted 17 lines · Yi L
Hi Peter,
Quoted 15 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 12:59 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContextQuoted 8 lines · Yi L
Quoted 4 lines
I'm not very clear on the relationship betwen an IOMMUContext and a DualStageIOMMUObject. Can there be many IOMMUContexts to a DualStageIOMMUOBject? The other way around? Or is it just zero-or-one DualStageIOMMUObjects to an IOMMUContext?
It is possible. As the below patch shows, DualStageIOMMUObject is per vfio container. IOMMUContext can be either per-device or shared across devices, it depends on vendor specific vIOMMU emulators.
Quoted 6 lines
Quoted 1 line
Is there an example when an IOMMUContext can be not per-device?
No, I don’t have such example so far. But as IOMMUContext is got from
pci_device_iommu_context(), in concept it possible to be not per-device. It is kind of leave to vIOMMU to decide if different devices could share a single IOMMUContext.
On the "pseries" machine the vIOMMU only has one set of translations for a whole virtual PCI Host Bridge (vPHB). So if you attach multiple devices to a single vPHB, I believe you'd get multiple devices in an IOMMUContext. Well.. if we did the PASID stuff, which we don't at the moment.
Note that on pseries on the other hand it's routine to create multiple vPHBs, rather than multiple PCI roots being an oddity as it is on x86.
Signature · 4 lines
David Gibson | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_ | _way_ _around_! http://www.ozlabs.org/~dgibson
RE: [RFC v3 03/25] hw/iommu: introduce IOMMUContext
Quoted 30 lines
From: David Gibson < [email protected] >
Sent: Friday, February 14, 2020 1:36 PM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContextQuoted 17 lines · Yi L
Hi Peter,
Quoted 15 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 12:59 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 03/25] hw/iommu: introduce IOMMUContextQuoted 8 lines · Yi L
Quoted 4 lines
I'm not very clear on the relationship betwen an IOMMUContext and a DualStageIOMMUObject. Can there be many IOMMUContexts to a DualStageIOMMUOBject? The other way around? Or is it just zero-or-one DualStageIOMMUObjects to an IOMMUContext?
It is possible. As the below patch shows, DualStageIOMMUObject is per vfio container. IOMMUContext can be either per-device or shared across devices, it depends on vendor specific vIOMMU emulators.
Quoted 6 lines
Quoted 1 line
Is there an example when an IOMMUContext can be not per-device?
No, I don’t have such example so far. But as IOMMUContext is got from
pci_device_iommu_context(), in concept it possible to be not per-device. It is kind of leave to vIOMMU to decide if different devices could share a single IOMMUContext.
Quoted 8 lines
On the "pseries" machine the vIOMMU only has one set of translations for a whole virtual PCI Host Bridge (vPHB). So if you attach multiple devices to a single vPHB, I believe you'd get multiple devices in an IOMMUContext. Well.. if we did the PASID stuff, which we don't at the moment.
Note that on pseries on the other hand it's routine to create multiple vPHBs, rather than multiple PCI roots being an oddity as it is on x86.
Thanks for the example, David. :-) BTW. I'll drop IOMMUContext in next version as the email below mentioned. Please feel free let me know your opinion.
https://lists.gnu.org/archive/html/qemu-devel/2020-02/msg02874.html
Regards,
Yi Liu
RFC v3 4/25 hw/pci: introduce pci_device_iommu_context()
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This patch adds pci_device_iommu_context() to get an iommu_context for a given device. A new callback is added in PCIIOMMUOps. Users who wants to call into vIOMMU could leaverage it. And further vIOMMU operation could be done via IOMMUContext.
Note:
Previous version has got Reviewed by from David Gibson and Peter Xu. But this version has slight changes due to the patch rebase. so I didn't add reviewed by from you two in case of that you two want to have one more review. If you two have no objection, I would add your review by in next version. Thanks.
--- hw/pci/pci.c | 35 ++++++++++++++++++++++++++++++----- include/hw/pci/pci.h | 5 +++++ 2 files changed, 35 insertions(+), 5 deletions(-)
hw/pci/pci.c +30 -5
@@ -2638,7 +2638,8 @@ 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;
@@ -2683,14 +2684,38 @@ 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; } +IOMMUContext *pci_device_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->get_iommu_context) { +return bus->iommu_ops->get_iommu_context (bus, + bus->iommu_opaque, devfn); + } +return NULL ; +} +void pci_setup_iommu (PCIBus *bus,const PCIIOMMUOps *ops,void *opaque) { bus->iommu_ops = ops;
include/hw/pci/pci.h +5 -0
@@ -9,6 +9,8 @@#include "hw/pci/pcie.h" +#include "hw/iommu/iommu_context.h" +extern bool pci_available;/* PCI bus */
@@ -488,9 +490,12 @@ typedef struct PCIIOMMUOps PCIIOMMUOps;struct PCIIOMMUOps {AddressSpace * (*get_address_space)(PCIBus *bus,void *opaque,int32_t devfn); +IOMMUContext * (*get_iommu_context)(PCIBus *bus, +void *opaque,int32_t devfn); };AddressSpace *pci_device_iommu_address_space (PCIDevice *dev); +IOMMUContext *pci_device_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 v3 5/25 intel_iommu: provide get_iommu_context() callback
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This patch adds get_iommu_context() callback to return an iommu_context Intel VT-d platform.
--- hw/i386/intel_iommu.c | 92 ++++++++++++++++++++++++++++++++++++++++--- include/hw/i386/intel_iommu.h | 15 ++++++- 2 files changed, 100 insertions(+), 7 deletions(-)
hw/i386/intel_iommu.c +86 -6
@@ -3347,22 +3347,35 @@ static const MemoryRegionOps vtd_mem_ir_ops = { }, }; -VTDAddressSpace *vtd_find_add_as (IntelIOMMUState *s,PCIBus *bus,int devfn) +/** +* Caller should hold iommu_lock. +*/ +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 ]; +VTDBus *vtd_bus; + vtd_bus =g_hash_table_lookup (s->vtd_as_by_busptr, &key);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) +PCI_DEVFN_MAX * \ + (sizeof (VTDAddressSpace *) +sizeof (VTDIOMMUContext *))); 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_iommu_lock (s); + vtd_bus =vtd_find_add_bus (s, bus); vtd_dev_as = vtd_bus->dev_as[devfn];
@@ -3426,9 +3439,63 @@ VTDAddressSpace *vtd_find_add_as(IntelIOMMUState *s, PCIBus *bus, int devfn)vtd_switch_address_space (vtd_dev_as); } +vtd_iommu_unlock (s); +return vtd_dev_as; } +static int vtd_icx_register_ds_iommu (IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj) +{ +VTDIOMMUContext *vtd_dev_icx =container_of (iommu_ctx, + VTDIOMMUContext, + iommu_context); + + vtd_dev_icx->dsi_obj = dsi_obj; +return 0 ; +} + +static void vtd_icx_unregister_ds_iommu (IOMMUContext *iommu_ctx, +DualStageIOMMUObject *dsi_obj) +{ +VTDIOMMUContext *vtd_dev_icx =container_of (iommu_ctx, + VTDIOMMUContext, + iommu_context); + + vtd_dev_icx->dsi_obj =NULL ; +} + +IOMMUContextOps vtd_iommu_context_ops = { + .register_ds_iommu = vtd_icx_register_ds_iommu, + .unregister_ds_iommu = vtd_icx_unregister_ds_iommu, +}; + +VTDIOMMUContext *vtd_find_add_icx (IntelIOMMUState *s, +PCIBus *bus,int devfn) +{ +VTDBus *vtd_bus; +VTDIOMMUContext *vtd_dev_icx; + +vtd_iommu_lock (s); + vtd_bus =vtd_find_add_bus (s, bus); + + vtd_dev_icx = vtd_bus->dev_icx[devfn]; + +if (!vtd_dev_icx) { + vtd_bus->dev_icx[devfn] = vtd_dev_icx = +g_malloc0 (sizeof (VTDIOMMUContext)); + vtd_dev_icx->vtd_bus = vtd_bus; + vtd_dev_icx->devfn = (uint8_t )devfn; + vtd_dev_icx->iommu_state = s; + vtd_dev_icx->dsi_obj =NULL ; +iommu_context_init (&vtd_dev_icx->iommu_context, + &vtd_iommu_context_ops); + } +vtd_iommu_unlock (s); + +return vtd_dev_icx; +} +static uint64_t get_naturally_aligned_size (uint64_t start,uint64_t size,int gaw) {
@@ -3722,8 +3789,21 @@ static AddressSpace *vtd_host_dma_iommu(PCIBus *bus, void *opaque, int devfn)return &vtd_as->as; } +static IOMMUContext *vtd_dev_iommu_context (PCIBus *bus, +void *opaque,int devfn) +{ +IntelIOMMUState *s = opaque; +VTDIOMMUContext *vtd_icx; + +assert (0 <= devfn && devfn <PCI_DEVFN_MAX ); + + vtd_icx =vtd_find_add_icx (s, bus, devfn); +return &vtd_icx->iommu_context; +} +static PCIIOMMUOps vtd_iommu_ops = { .get_address_space = vtd_host_dma_iommu, + .get_iommu_context = vtd_dev_iommu_context, };static bool vtd_decide_config (IntelIOMMUState *s,Error **errp)
include/hw/i386/intel_iommu.h +14 -1
@@ -68,6 +68,7 @@ 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 VTDIOMMUContext VTDIOMMUContext ;/* Context-Entry */ struct VTDContextEntry {
@@ -116,9 +117,20 @@ struct VTDAddressSpace {IOVATree *iova_tree;/* Traces mapped IOVA ranges */ }; +struct VTDIOMMUContext { +VTDBus *vtd_bus; +uint8_t devfn; +IOMMUContext iommu_context; +DualStageIOMMUObject *dsi_obj; +IntelIOMMUState *iommu_state; +}; +struct VTDBus {PCIBus * bus;/* A reference to the bus to provide translation for */ -VTDAddressSpace *dev_as[0 ];/* A table of VTDAddressSpace objects indexed by devfn */ +/* A table of VTDAddressSpace objects indexed by devfn */ +VTDAddressSpace *dev_as[PCI_DEVFN_MAX ]; +/* A table of VTDIOMMUContext objects indexed by devfn */ +VTDIOMMUContext *dev_icx[PCI_DEVFN_MAX ]; };struct VTDIOTLBEntry {
@@ -282,5 +294,6 @@ struct IntelIOMMUState {* create a new one if none exists */ VTDAddressSpace *vtd_find_add_as (IntelIOMMUState *s,PCIBus *bus,int devfn); +VTDIOMMUContext *vtd_find_add_icx (IntelIOMMUState *s,PCIBus *bus,int devfn);#endif
Signature · 1 line
2.7.4
RFC v3 6/25 scripts/update-linux-headers: Import iommu.h
In reply to Liu, Yi LFrom: 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
Re: [RFC v3 06/25] scripts/update-linux-headers: Import iommu.h
Quoted 21 lines · Yi L
From: Eric Auger <[email protected]>
Update the script to import the new
iommu.huapi header.Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Michael S. Tsirkin <[email protected]>Cc Cornelia Huck <[email protected]>Cc Paolo Bonzini <[email protected]>Signed-off-by Eric Auger <[email protected]>--- scripts/update-linux-headers.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)scripts/update-linux-headers.sh +0
-0
scripts/update-linux-headers.sh done · line 147
Quoted 8 lines
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" "$output/linux-headers/linux" done
RE: [RFC v3 06/25] scripts/update-linux-headers: Import iommu.h
Quoted 30 lines
From: Cornelia Huck [mailto:[email protected]]
Sent: Wednesday, January 29, 2020 8:25 PM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 06/25]scripts/update-linux-headers: Importiommu.hQuoted 22 lines · Yi L
From: Eric Auger <[email protected]>
Update the script to import the new
iommu.huapi header.Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Michael S. Tsirkin <[email protected]>Cc Cornelia Huck <[email protected]>Cc Paolo Bonzini <[email protected]>Signed-off-by Eric Auger <[email protected]>--- scripts/update-linux-headers.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)scripts/update-linux-headers.sh +0
-0@@ -141,7 +141,7 @@ done
Quoted 9 lines
scripts/update-linux-headers.sh done · line 147
Quoted 7 lines
scripts/update-linux-headers.sh +1
-1rm -rf "$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" doneAcked-by Cornelia Huck <[email protected]>
Thanks, Cornelia.
Regards,
Yi Liu
RFC v3 7/25 header file update VFIO/IOMMU vSVA APIs
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
The 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.
--- linux-headers/linux/iommu.h | 372 ++++++++++++++++++++++++++++++++++++++++++++ linux-headers/linux/vfio.h | 148 ++++++++++++++++++ 2 files changed, 520 insertions(+) create mode 100644 linux-headers/linux/iommu.h
linux-headers/linux/iommu.h new file +372 -0
@@ -0,0 +1,372 @@ +/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ +/* +* IOMMU user API definitions +*/ + +#ifndef _IOMMU_H +#define _IOMMU_H + +#include <linux/types.h> + +/** +* Current version of the IOMMU user API. This is intended for query +* between user and kernel to determine compatible data structures. +* +* Having a single UAPI version to govern the user-kernel data structures +* makes compatibility check straightforward. On the contrary, supporting +* combinations of multiple versions of the data can be a nightmare. +* +* UAPI version can be bumped up with the following rules: +* 1. All data structures passed between user and kernel space share +* the same version number. i.e. any extension to to any structure +* results in version bump up. +* +* 2. Data structures are open to extension but closed to modification. +* New fields must be added at the end of each data structure with +* 64bit alignment. Flag bits can be added without size change but +* existing ones cannot be altered. +* +* 3. Versions are backward compatible. +* +* 4. Version to size lookup is supported by kernel internal API for each +* API function type. @version is mandatory for new data structures +* and must be at the beginning with type of __u32. +*/ +#define IOMMU_UAPI_VERSION 1 +static __inline__int iommu_get_uapi_version (void ) +{ +return IOMMU_UAPI_VERSION ; +} + +/* +* Supported UAPI features that can be reported to user space. +* These types represent the capability available in the kernel. +* +* REVISIT: UAPI version also implies the capabilities. Should we +* report them explicitly? +*/ +enum IOMMU_UAPI_DATA_TYPES { +IOMMU_UAPI_BIND_GPASID , +IOMMU_UAPI_CACHE_INVAL , +IOMMU_UAPI_PAGE_RESP , +NR_IOMMU_UAPI_TYPE , +}; + +#define IOMMU_UAPI_CAP_MASK ((1 << IOMMU_UAPI_BIND_GPASID) | \ + (1 << IOMMU_UAPI_CACHE_INVAL) | \ + (1 << IOMMU_UAPI_PAGE_RESP)) + +#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) +* @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) +__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 +* @version: IOMMU_UAPI_VERSION +* @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 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 +* @version: IOMMU_UAPI_VERSION +* @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 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[2 ]; +union { +struct iommu_inv_pasid_info pasid_info; +struct iommu_inv_addr_info addr_info; + }; +}; + +/** +* 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 */ +__u64 flags; +__u32 pat; +__u32 emt; +}; +#define IOMMU_SVA_VTD_GPASID_EMT_MASK (IOMMU_SVA_VTD_GPASID_CD | \ + IOMMU_SVA_VTD_GPASID_EMTE | \ + IOMMU_SVA_VTD_GPASID_PCD | \ + IOMMU_SVA_VTD_GPASID_PWT) +/** +* struct iommu_gpasid_bind_data - Information about device and guest PASID binding +* @version: IOMMU_UAPI_VERSION +* @format: PASID table entry format +* @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 version; +#define IOMMU_PASID_FORMAT_INTEL_VTD 1 +__u32 format; +#define IOMMU_SVA_GPASID_VAL (1 << 0)/* guest PASID valid */ +__u64 flags; +__u64 gpgd; +__u64 hpasid; +__u64 gpasid; +__u32 addr_width; +__u8 padding[12 ]; +/* Vendor specific data */ +union { +struct iommu_gpasid_bind_data_vtd vtd; + }; +}; + +#endif /* _IOMMU_H */
linux-headers/linux/vfio.h +148 -0
166 lines not on this page · raw message
Signature · 1 line
2.7.4
Re: [RFC v3 07/25] header file update VFIO/IOMMU vSVA APIs
Quoted 14 lines · Yi L
From: Liu Yi L <[email protected]>
The kernel
uapi/linux/iommu.hheader file includes the extensions for vSVA support. e.g. bind gpasid, iommu fault report related user structures and etc.Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Michael S. Tsirkin <[email protected]>Cc Cornelia Huck <[email protected]>Cc Paolo Bonzini <[email protected]>Signed-off-by Liu Yi L <[email protected]>
Quoted 5 lines
--- linux-headers/linux/iommu.h | 372 ++++++++++++++++++++++++++++++++++++++++++++ linux-headers/linux/vfio.h | 148 ++++++++++++++++++ 2 files changed, 520 insertions(+) create mode 100644 linux-headers/linux/iommu.h
Please add a note that this is to be replaced with a full headers update, so that it doesn't get missed :)
RE: [RFC v3 07/25] header file update VFIO/IOMMU vSVA APIs
Quoted 30 lines
From: [email protected] [mailto:[email protected]] On Behalf Of Cornelia Huck
Sent: Wednesday, January 29, 2020 8:29 PM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 07/25] header file update VFIO/IOMMU vSVA APIsQuoted 14 lines · Yi L
From: Liu Yi L <[email protected]>
The kernel
uapi/linux/iommu.hheader file includes the extensions for vSVA support. e.g. bind gpasid, iommu fault report related user structures and etc.Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Michael S. Tsirkin <[email protected]>Cc Cornelia Huck <[email protected]>Cc Paolo Bonzini <[email protected]>Signed-off-by Liu Yi L <[email protected]>Quoted 6 lines
--- linux-headers/linux/iommu.h | 372 ++++++++++++++++++++++++++++++++++++++++++++ linux-headers/linux/vfio.h | 148 ++++++++++++++++++ 2 files changed, 520 insertions(+) create mode 100644 linux-headers/linux/iommu.h
Quoted 2 lines
Please add a note that this is to be replaced with a full headers update, so that it doesn't get missed :)
Exactly, thanks for the reminder. I expect to have a full headers update when the whole vSVA series is accepted. :-)
Thanks,
Yi Liu
RFC v3 8/25 vfio: pass IOMMUContext into vfio_get_group()
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
IOMMUContext provides an explicit way for VFIO to talk with vIOMMU.
This patch passes IOMMUContext instance into vfio_get_group() for the reason of potential VFIO_TYPE1_NESTING_IOMMU configuration as such configuration requires interaction with vIOMMU.
--- hw/vfio/ap.c | 2 +- hw/vfio/ccw.c | 2 +- hw/vfio/common.c | 3 ++- hw/vfio/pci.c | 3 ++- hw/vfio/platform.c | 2 +- include/hw/vfio/vfio-common.h | 4 +++- 6 files changed, 10 insertions(+), 6 deletions(-)
hw/vfio/ap.c +1 -1
@@ -82,7 +82,7 @@ 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,NULL , errp); }static void vfio_ap_realize (DeviceState *dev,Error **errp)
hw/vfio/ccw.c +1 -1
@@ -471,7 +471,7 @@ 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,NULL , errp); }static void vfio_ccw_realize (DeviceState *dev,Error **errp)
hw/vfio/common.c +2 -1
@@ -1453,7 +1453,8 @@ 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, +IOMMUContext *iommu_ctx,Error **errp) {VFIOGroup *group;char path[32 ];
hw/vfio/pci.c +2 -1
@@ -2782,7 +2782,8 @@ 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); + group =vfio_get_group (groupid,pci_device_iommu_address_space (pdev), +pci_device_iommu_context (pdev), errp);if (!group) { goto error; }
hw/vfio/platform.c +1 -1
@@ -581,7 +581,7 @@ 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,NULL , errp);if (!group) {return -ENOENT ; }
include/hw/vfio/vfio-common.h +3 -1
@@ -26,6 +26,7 @@#include "qemu/notify.h" #include "ui/console.h" #include "hw/display/ramfb.h" +#include "hw/iommu/iommu_context.h" #ifdef CONFIG_LINUX #include <linux/vfio.h> #endif
@@ -174,7 +175,8 @@ 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, +IOMMUContext *iommu_ctx,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 v3 9/25 vfio: check VFIO_TYPE1_NESTING_IOMMU support
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
VFIO needs to check VFIO_TYPE1_NESTING_IOMMU support with Kernel before further using it. e.g. requires to check IOMMU UAPI version.
--- hw/vfio/common.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)
hw/vfio/common.c +12 -2
@@ -1157,12 +1157,21 @@ static void vfio_put_address_space(VFIOAddressSpace *space)static int vfio_get_iommu_type (VFIOContainer *container,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, version;for (i =0 ; i <ARRAY_SIZE (iommu_types); i++) {if (ioctl (container->fd,VFIO_CHECK_EXTENSION , iommu_types[i])) { +if (iommu_types[i] ==VFIO_TYPE1_NESTING_IOMMU ) { + version =ioctl (container->fd, +VFIO_NESTING_GET_IOMMU_UAPI_VERSION ); +if (version <IOMMU_UAPI_VERSION ) { +printf ("IOMMU UAPI incompatible for nesting\n" ); +continue ; + } + }return iommu_types[i]; } }
@@ -1278,6 +1287,7 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as, }switch (container->iommu_type) { +case VFIO_TYPE1_NESTING_IOMMU :case VFIO_TYPE1v2_IOMMU:case VFIO_TYPE1_IOMMU : {
Signature · 1 line
2.7.4
Re: [RFC v3 09/25] vfio: check VFIO_TYPE1_NESTING_IOMMU support
Quoted 31 lines · Yi L
From: Liu Yi L <[email protected]>
VFIO needs to check
VFIO_TYPE1_NESTING_IOMMUsupport with Kernel before further using it. e.g. requires to check IOMMU UAPI version.Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Cc Alex Williamson <[email protected]>Signed-off-by Liu Yi L <[email protected]>Signed-off-by Yi Sun <[email protected]>--- hw/vfio/common.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)hw/vfio/common.c +2
-1@@ -1157,12 +1157,21 @@ static void vfio_put_address_space(VFIOAddressSpace *space)static int vfio_get_iommu_type (VFIOContainer *container,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 };
hw/vfio/common.c static void vfio_put_address_space(VFIOAddressSpace *space) · line 1171
Quoted 10 lines
hw/vfio/common.c +6
-1-int i; +int i, version;for (i =0 ; i <ARRAY_SIZE (iommu_types); i++) {if (ioctl (container->fd,VFIO_CHECK_EXTENSION , iommu_types[i])) { +if (iommu_types[i] ==VFIO_TYPE1_NESTING_IOMMU ) { + version =ioctl (container->fd, +VFIO_NESTING_GET_IOMMU_UAPI_VERSION ); +if (version <IOMMU_UAPI_VERSION ) { +printf ("IOMMU UAPI incompatible for nesting\n" );
There should have better alternatives than printf()... Maybe warn_report()?
Quoted 17 lines
hw/vfio/common.c +4
-0+continue ; + } + }return iommu_types[i]; } }@@ -1278,6 +1287,7 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as, }switch (container->iommu_type) { +case VFIO_TYPE1_NESTING_IOMMU :case VFIO_TYPE1v2_IOMMU:case VFIO_TYPE1_IOMMU : {Signature · 1 line
2.7.4
Signature · 1 line
Peter Xu
RE: [RFC v3 09/25] vfio: check VFIO_TYPE1_NESTING_IOMMU support
Quoted 48 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 3:08 AM
Subject: Re: [RFC v3 09/25] vfio: checkVFIO_TYPE1_NESTING_IOMMUsupportQuoted 24 lines · Yi L
From: Liu Yi L <[email protected]>
VFIO needs to check
VFIO_TYPE1_NESTING_IOMMUsupport with Kernel before further using it.
e.g. requires to check IOMMU UAPI version.Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Cc Alex Williamson <[email protected]>Signed-off-by Liu Yi L <[email protected]>Signed-off-by Yi Sun <[email protected]>--- hw/vfio/common.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-)hw/vfio/common.c index renamed from hw/vfio/common.c +0
-00cc7ff5..a5e70b1 100644
hw/vfio/common.c +0
-0@@ -1157,12 +1157,21 @@ static voidQuoted 1 line
vfio_put_address_space(VFIOAddressSpace *space)static intvfio_get_iommu_type(VFIOContainer *container,
Quoted 11 lines
Error **errp) {hw/vfio/common.c +3
-2-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, version;for (i =0 ; i <ARRAY_SIZE (iommu_types); i++) { if (ioctl(container->fd, VFIO_CHECK_EXTENSION,iommu_types[i])) {
hw/vfio/common.c static void vfio_put_address_space(VFIOAddressSpace *space) · line 1171
Quoted 5 lines
hw/vfio/common.c +5
-0+if (iommu_types[i] ==VFIO_TYPE1_NESTING_IOMMU ) { + version =ioctl (container->fd, +VFIO_NESTING_GET_IOMMU_UAPI_VERSION ); +if (version <IOMMU_UAPI_VERSION ) { +printf ("IOMMU UAPI incompatible for nesting\n" );
Quoted 1 line
There should have better alternatives than
printf()... Maybewarn_report()?
Got it. thanks. 😊
Regards,
Yi Liu
RFC v3 10/25 vfio: register DualStageIOMMUObject to vIOMMU
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
After confirming dual stage DMA translation support with kernel by checking VFIO_TYPE1_NESTING_IOMMU, VFIO could register DualStageIOMMUObject instance to vIOMMU, thus that vIOMMU is aware of such hardware capability. vIOMMU may make use of such capability by leveraging the ops provided by DualStageIOMMUObject.
--- hw/vfio/common.c | 30 ++++++++++++++++++++++++++++-- include/hw/vfio/vfio-common.h | 2 ++ 2 files changed, 30 insertions(+), 2 deletions(-)
hw/vfio/common.c +28 -2
@@ -1179,6 +1179,9 @@ static int vfio_get_iommu_type(VFIOContainer *container,return -EINVAL ; } +static struct DualStageIOMMUOps vfio_ds_iommu_ops = { +}; +static int vfio_init_container (VFIOContainer *container,int group_fd,Error **errp) {
@@ -1210,12 +1213,29 @@ static int vfio_init_container(VFIOContainer *container, int group_fd,return -errno; } +if (iommu_type ==VFIO_TYPE1_NESTING_IOMMU ) { +ds_iommu_object_init (&container->dsi_obj, &vfio_ds_iommu_ops); +if (iommu_context_register_ds_iommu (container->iommu_ctx, + &container->dsi_obj)) { +/* +* Here just need an info to indicate that there is no +* DualStageIOMMUObject instance registered to vIOMMU +* due to either no IOMMUContext support in vIOMMU or +* vIOMMU internal failure. Neither is fatal error to +* VFIO as it is not mandatory requirement to use such +* capability in vIOMMU. +*/ +printf ("No Dual Stage IOMMU for container(0x%p)\n" , container); +ds_iommu_object_destroy (&container->dsi_obj); + } + } + container->iommu_type = iommu_type;return 0 ; }static int vfio_connect_container (VFIOGroup *group,AddressSpace *as, -Error **errp) +IOMMUContext *iommu_ctx,Error **errp) {VFIOContainer *container;int ret, fd;
@@ -1277,6 +1297,7 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as, container =g_malloc0 (sizeof (*container)); container->space = space; container->fd = fd; + container->iommu_ctx = iommu_ctx; container->error =NULL ;QLIST_INIT (&container->giommu_list);QLIST_INIT (&container->hostwin_list);
@@ -1457,6 +1478,11 @@ static void vfio_disconnect_container(VFIOGroup *group)trace_vfio_disconnect_container (container->fd);close (container->fd); +if (container->iommu_ctx) { +iommu_context_unregister_ds_iommu (container->iommu_ctx, + &container->dsi_obj); + } +ds_iommu_object_destroy (&container->dsi_obj);g_free (container);vfio_put_address_space (space);
@@ -1508,7 +1534,7 @@ 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, iommu_ctx, errp)) {error_prepend (errp,"failed to setup container for group %d: " , groupid); goto close_fd_exit;
include/hw/vfio/vfio-common.h +2 -0
@@ -72,6 +72,8 @@ typedef struct VFIOContainer {MemoryListener listener;MemoryListener prereg_listener;unsigned iommu_type; +IOMMUContext *iommu_ctx; +DualStageIOMMUObject dsi_obj;Error *error;bool initialized;unsigned long pgsizes;
Signature · 1 line
2.7.4
RFC v3 11/25 vfio: get stage-1 pasid formats from Kernel
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
VFIO checks IOMMU UAPI version when it finds Kernel supports VFIO_TYPE1_NESTING_IOMMU. It is enough for UAPI compatibility check. However, IOMMU UAPI may support multiple stage-1 pasid formats in a specific UAPI version, which is highly possible since IOMMU UAPI supports stage-1 formats across all IOMMU vendors. So VFIO needs to get the supported formats from Kernel and tell vIOMMU. Let vIOMMU select proper format when setup dual stage DMA translation.
This patch gets the stage-1 pasid format from kernel by using IOCTL VFIO_IOMMU_GET_INFO and pass the supported format to vIOMMU by the DualStageIOMMUObject instance which has been registered to vIOMMU.
This patch referred some code from Shameer Kolothum.
https://lists.gnu.org/archive/html/qemu-devel/2018-05/msg03759.html
--- hw/iommu/dual_stage_iommu.c | 5 ++- hw/vfio/common.c | 85 ++++++++++++++++++++++++++++++++++++- include/hw/iommu/dual_stage_iommu.h | 10 ++++- 3 files changed, 97 insertions(+), 3 deletions(-)
hw/iommu/dual_stage_iommu.c +4 -1
@@ -48,9 +48,12 @@ int ds_iommu_pasid_free(DualStageIOMMUObject *dsi_obj, uint32_t pasid) }void ds_iommu_object_init (DualStageIOMMUObject *dsi_obj, -DualStageIOMMUOps *ops) +DualStageIOMMUOps *ops, +DualStageIOMMUInfo *uinfo) { dsi_obj->ops = ops; + + dsi_obj->uinfo.pasid_format = uinfo->pasid_format; }void ds_iommu_object_destroy (DualStageIOMMUObject *dsi_obj)
hw/vfio/common.c +84 -1
101 lines not on this page · raw message
include/hw/iommu/dual_stage_iommu.h +9 -1
@@ -23,12 +23,14 @@#define HW_DS_IOMMU_H #include "qemu/queue.h" +#include <linux/iommu.h> #ifndef CONFIG_USER_ONLY #include "exec/hwaddr.h" #endif typedef struct DualStageIOMMUObject DualStageIOMMUObject ;typedef struct DualStageIOMMUOps DualStageIOMMUOps ; +typedef struct DualStageIOMMUInfo DualStageIOMMUInfo ;struct DualStageIOMMUOps {/* Allocate pasid from DualStageIOMMU (a.k.a. host IOMMU) */
@@ -41,11 +43,16 @@ struct DualStageIOMMUOps { uint32_t pasid); }; +struct DualStageIOMMUInfo { +uint32_t pasid_format; +}; +/* * This is an abstraction of Dual-stage IOMMU. */ struct DualStageIOMMUObject {DualStageIOMMUOps *ops; +DualStageIOMMUInfo uinfo; }; int ds_iommu_pasid_alloc(DualStageIOMMUObject *dsi_obj, uint32_t min,
@@ -53,7 +60,8 @@ int ds_iommu_pasid_alloc(DualStageIOMMUObject *dsi_obj, uint32_t min,int ds_iommu_pasid_free (DualStageIOMMUObject *dsi_obj,uint32_t pasid);void ds_iommu_object_init (DualStageIOMMUObject *dsi_obj, -DualStageIOMMUOps *ops); +DualStageIOMMUOps *ops, +DualStageIOMMUInfo *uinfo);void ds_iommu_object_destroy (DualStageIOMMUObject *dsi_obj);#endif
Signature · 1 line
2.7.4
Re: [RFC v3 11/25] vfio: get stage-1 pasid formats from Kernel
Quoted 56 lines · Yi L
From: Liu Yi L <[email protected]>
VFIO checks IOMMU UAPI version when it finds Kernel supports
VFIO_TYPE1_NESTING_IOMMU. It is enough for UAPI compatibility check. However, IOMMU UAPI may support multiple stage-1 pasid formats in a specific UAPI version, which is highly possible since IOMMU UAPI supports stage-1 formats across all IOMMU vendors. So VFIO needs to get the supported formats from Kernel and tell vIOMMU. Let vIOMMU select proper format when setup dual stage DMA translation.This patch gets the stage-1 pasid format from kernel by using IOCTL
VFIO_IOMMU_GET_INFOand pass the supported format to vIOMMU by the DualStageIOMMUObject instance which has been registered to vIOMMU.This patch referred some code from Shameer Kolothum.
https://lists.gnu.org/archive/html/qemu-devel/2018-05/msg03759.htmlCc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Cc Alex Williamson <[email protected]>Signed-off-by Shameer Kolothum <[email protected]>Signed-off-by Liu Yi L <[email protected]>--- hw/iommu/dual_stage_iommu.c | 5 ++- hw/vfio/common.c | 85 ++++++++++++++++++++++++++++++++++++- include/hw/iommu/dual_stage_iommu.h | 10 ++++- 3 files changed, 97 insertions(+), 3 deletions(-)hw/iommu/dual_stage_iommu.c +4
-1@@ -48,9 +48,12 @@ int ds_iommu_pasid_free(DualStageIOMMUObject *dsi_obj, uint32_t pasid) }void ds_iommu_object_init (DualStageIOMMUObject *dsi_obj, -DualStageIOMMUOps *ops) +DualStageIOMMUOps *ops, +DualStageIOMMUInfo *uinfo) { dsi_obj->ops = ops; + + dsi_obj->uinfo.pasid_format = uinfo->pasid_format; }void ds_iommu_object_destroy (DualStageIOMMUObject *dsi_obj)hw/vfio/common.c +0
-0
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1186
Quoted 5 lines
hw/vfio/common.c +2
-0@@ -1182,10 +1182,84 @@ static int vfio_get_iommu_type(VFIOContainer *container,static struct DualStageIOMMUOps vfio_ds_iommu_ops = { }; +static int vfio_get_iommu_info (VFIOContainer *container, +struct vfio_iommu_type1_info **info)
Better comment on the function to remember to free(*info) after use for the callers.
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1190
Quoted 4 lines
hw/vfio/common.c +4
-0+{ + +size_t argsz =sizeof (struct vfio_iommu_type1_info ); +
Nit: extra newline.
Quoted 39 lines
hw/vfio/common.c +39
-0+ + *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 ; +}
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1239
Quoted 10 lines
hw/vfio/common.c +10
-0+ +static int vfio_get_nesting_iommu_format (VFIOContainer *container, +uint32_t *pasid_format) +{ +struct vfio_iommu_type1_info *info; +struct vfio_info_cap_header *hdr; +struct vfio_iommu_type1_info_cap_nesting *cap; + +if (vfio_get_iommu_info (container, &info)) { +return -errno;
Should return the retcode from vfio_get_iommu_info.
Quoted 27 lines
hw/vfio/common.c +19
-0+ } + + hdr =vfio_get_iommu_info_cap (info, +VFIO_IOMMU_TYPE1_INFO_CAP_NESTING ); +if (!hdr) { +g_free (info); +return -errno; + } + + cap =container_of (hdr, + struct vfio_iommu_type1_info_cap_nesting, header); + *pasid_format = cap->pasid_format; + +g_free (info); +return 0 ; +} +static int vfio_init_container (VFIOContainer *container,int group_fd,Error **errp) {int iommu_type, ret; +uint32_t format; +DualStageIOMMUInfo uinfo; iommu_type =vfio_get_iommu_type (container, errp);if (iommu_type <0 ) {
hw/vfio/common.c static int vfio_init_container(VFIOContainer *container, int group_fd, · line 1294
Quoted 8 lines
hw/vfio/common.c +4
-1@@ -1214,7 +1288,16 @@ static int vfio_init_container(VFIOContainer *container, int group_fd, }if (iommu_type ==VFIO_TYPE1_NESTING_IOMMU ) { -ds_iommu_object_init (&container->dsi_obj, &vfio_ds_iommu_ops); +if (vfio_get_nesting_iommu_format (container, &format)) { +error_setg_errno (errp, errno, +"Failed to get nesting iommu format" ); +return -errno;
Same here, you might want to return the retcode from vfio_get_nesting_iommu_format()?
Quoted 58 lines
hw/vfio/common.c +6
-0+ } + + uinfo.pasid_format = format; +ds_iommu_object_init (&container->dsi_obj, + &vfio_ds_iommu_ops, &uinfo); +if (iommu_context_register_ds_iommu (container->iommu_ctx, &container->dsi_obj)) {/* include/hw/iommu/dual_stage_iommu.h +9
-1@@ -23,12 +23,14 @@#define HW_DS_IOMMU_H #include "qemu/queue.h" +#include <linux/iommu.h> #ifndef CONFIG_USER_ONLY #include "exec/hwaddr.h" #endif typedef struct DualStageIOMMUObject DualStageIOMMUObject ;typedef struct DualStageIOMMUOps DualStageIOMMUOps ; +typedef struct DualStageIOMMUInfo DualStageIOMMUInfo ;struct DualStageIOMMUOps {/* Allocate pasid from DualStageIOMMU (a.k.a. host IOMMU) */ @@ -41,11 +43,16 @@ struct DualStageIOMMUOps { uint32_t pasid); }; +struct DualStageIOMMUInfo { +uint32_t pasid_format; +}; +/* * This is an abstraction of Dual-stage IOMMU. */ struct DualStageIOMMUObject {DualStageIOMMUOps *ops; +DualStageIOMMUInfo uinfo; }; int ds_iommu_pasid_alloc(DualStageIOMMUObject *dsi_obj, uint32_t min,@@ -53,7 +60,8 @@ int ds_iommu_pasid_alloc(DualStageIOMMUObject *dsi_obj, uint32_t min,int ds_iommu_pasid_free (DualStageIOMMUObject *dsi_obj,uint32_t pasid);void ds_iommu_object_init (DualStageIOMMUObject *dsi_obj, -DualStageIOMMUOps *ops); +DualStageIOMMUOps *ops, +DualStageIOMMUInfo *uinfo);void ds_iommu_object_destroy (DualStageIOMMUObject *dsi_obj);#endif Signature · 1 line
2.7.4
Signature · 1 line
Peter Xu
RE: [RFC v3 11/25] vfio: get stage-1 pasid formats from Kernel
Quoted 66 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 3:30 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 11/25] vfio: get stage-1 pasid formats from KernelQuoted 27 lines · Yi L
From: Liu Yi L <[email protected]>
VFIO checks IOMMU UAPI version when it finds Kernel supports
VFIO_TYPE1_NESTING_IOMMU. It is enough for UAPI compatibility check. However, IOMMU UAPI may support multiple stage-1 pasid formats in a specific UAPI version, which is highly possible since IOMMU UAPI supports stage-1 formats across all IOMMU vendors.
So VFIO needs to get the supported formats from Kernel and tell vIOMMU. Let vIOMMU select proper format when setup dual stage DMA translation.This patch gets the stage-1 pasid format from kernel by using IOCTL
VFIO_IOMMU_GET_INFOand pass the supported format to vIOMMU by the DualStageIOMMUObject instance which has been registered to vIOMMU.This patch referred some code from Shameer Kolothum.
https://lists.gnu.org/archive/html/qemu-devel/2018-05/msg03759.htmlCc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Cc Alex Williamson <[email protected]>Signed-off-by Shameer Kolothum <[email protected]>Signed-off-by Liu Yi L <[email protected]>Quoted 3 lines
--- hw/iommu/dual_stage_iommu.c | 5 ++- hw/vfio/common.c | 85++++++++++++++++++++++++++++++++++++-
Review
Quoted 4 lines
include/hw/iommu/dual_stage_iommu.h | 10 ++++- 3 files changed, 97 insertions(+), 3 deletions(-)(quoted diff) +0
-0
b/hw/iommu/dual_stage_iommu.cQuoted 24 lines
index be4179d..d5a7168 100644
hw/iommu/dual_stage_iommu.c +4
-2@@ -48,9 +48,12 @@ int ds_iommu_pasid_free(DualStageIOMMUObject *dsi_obj, uint32_t pasid) }void ds_iommu_object_init (DualStageIOMMUObject *dsi_obj, -DualStageIOMMUOps *ops) +DualStageIOMMUOps *ops, +DualStageIOMMUInfo *uinfo) { dsi_obj->ops = ops; + + dsi_obj->uinfo.pasid_format = uinfo->pasid_format; } void ds_iommu_object_destroy(DualStageIOMMUObject *dsi_obj) diff --git a/hw/vfio/common.c b/hw/vfio/common.c index fc1723d..a07824b100644
hw/vfio/common.c +0
-0@@ -1182,10 +1182,84 @@ static int vfio_get_iommu_type(VFIOContainer *container, static struct DualStageIOMMUOps vfio_ds_iommu_ops = { };
Quoted 5 lines
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1186
Quoted 2 lines
hw/vfio/common.c +2
-0+static int vfio_get_iommu_info (VFIOContainer *container, +struct vfio_iommu_type1_info **info)Better comment on the function to remember to
free(*info)after use for the callers.
Will do. 😊
Quoted 7 lines
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1190
Quoted 5 lines
hw/vfio/common.c +4
-0+{ + +size_t argsz =sizeof (struct vfio_iommu_type1_info ); +Nit: extra newline.
accepted. 😊
Quoted 42 lines
Quoted 42 lines
hw/vfio/common.c +42
-0+ + *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_format (VFIOContainer *container, +uint32_t *pasid_format) {
Quoted 8 lines
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1239
Quoted 6 lines
hw/vfio/common.c +6
-0+struct vfio_iommu_type1_info *info; +struct vfio_info_cap_header *hdr; +struct vfio_iommu_type1_info_cap_nesting *cap; + +if (vfio_get_iommu_info (container, &info)) { +return -errno;Should return the retcode from
vfio_get_iommu_info.
yes , it is. thx for catching it.
Quoted 36 lines
hw/vfio/common.c static int vfio_init_container(VFIOContainer
Quoted 26 lines
hw/vfio/common.c +19
-0+ } + + hdr =vfio_get_iommu_info_cap (info, +VFIO_IOMMU_TYPE1_INFO_CAP_NESTING ); +if (!hdr) { +g_free (info); +return -errno; + } + + cap =container_of (hdr, + struct vfio_iommu_type1_info_cap_nesting, header); + *pasid_format = cap->pasid_format; + +g_free (info); +return 0 ; +} +static int vfio_init_container (VFIOContainer *container,int group_fd,Error **errp) {int iommu_type, ret; +uint32_t format; +DualStageIOMMUInfo uinfo; iommu_type =vfio_get_iommu_type (container, errp);if (iommu_type <0 ) {@@ -1214,7 +1288,16 @@ static int vfio_init_container(VFIOContainer*container, int
group_fd,hw/vfio/common.c static int vfio_init_container(VFIOContainer *container, int group_fd, · line 1294
Quoted 8 lines
hw/vfio/common.c +4
-1}if (iommu_type ==VFIO_TYPE1_NESTING_IOMMU ) { -ds_iommu_object_init (&container->dsi_obj, &vfio_ds_iommu_ops); +if (vfio_get_nesting_iommu_format (container, &format)) { +error_setg_errno (errp, errno, +"Failed to get nesting iommu format" ); +return -errno;
Quoted 2 lines
Same here, you might want to return the retcode from
vfio_get_nesting_iommu_format()?
will do it. 😊
Thanks for your comments, I'll address them in next version.
Regards,
Yi Liu
RFC v3 12/25 vfio/common: add pasid_alloc/free support
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This patch adds VFIO pasid alloc/free support to allow host intercept in PASID allocation for VM by adding VFIO implementation of DualStageIOMMUOps.pasid_alloc/free callbacks.
--- hw/vfio/common.c | 42 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+)
hw/vfio/common.c +42 -0
@@ -1179,7 +1179,49 @@ static int vfio_get_iommu_type(VFIOContainer *container,return -EINVAL ; } +static int vfio_ds_iommu_pasid_alloc (DualStageIOMMUObject *dsi_obj, +uint32_t min,uint32_t max,uint32_t *pasid) +{ +VFIOContainer *container =container_of (dsi_obj, VFIOContainer, dsi_obj); +struct vfio_iommu_type1_pasid_request req; +unsigned long argsz; + + argsz =sizeof (req); + req.argsz = argsz; + req.flags =VFIO_IOMMU_PASID_ALLOC ; + req.alloc_pasid.min = min; + req.alloc_pasid.max = max; + +if (ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req)) { +error_report ("%s: %d, alloc failed" , __func__, -errno); +return -errno; + } + *pasid = req.alloc_pasid.result; +return 0 ; +} + +static int vfio_ds_iommu_pasid_free (DualStageIOMMUObject *dsi_obj, +uint32_t pasid) +{ +VFIOContainer *container =container_of (dsi_obj, VFIOContainer, dsi_obj); +struct vfio_iommu_type1_pasid_request req; +unsigned long argsz; + + argsz =sizeof (req); + req.argsz = argsz; + req.flags =VFIO_IOMMU_PASID_FREE ; + req.free_pasid = pasid; + +if (ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req)) { +error_report ("%s: %d, free failed" , __func__, -errno); +return -errno; + } +return 0 ; +} +static struct DualStageIOMMUOps vfio_ds_iommu_ops = { + .pasid_alloc = vfio_ds_iommu_pasid_alloc, + .pasid_free = vfio_ds_iommu_pasid_free, }; static int vfio_get_iommu_info(VFIOContainer *container,
Signature · 1 line
2.7.4
Re: [RFC v3 12/25] vfio/common: add pasid_alloc/free support
Quoted 32 lines · Yi L
From: Liu Yi L <[email protected]>
This patch adds VFIO pasid alloc/free support to allow host intercept in PASID allocation for VM by adding VFIO implementation of DualStageIOMMUOps.pasid_alloc/free callbacks.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Cc Alex Williamson <[email protected]>Signed-off-by Liu Yi L <[email protected]>--- hw/vfio/common.c | 42 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+)hw/vfio/common.c +6
-0@@ -1179,7 +1179,49 @@ static int vfio_get_iommu_type(VFIOContainer *container,return -EINVAL ; } +static int vfio_ds_iommu_pasid_alloc (DualStageIOMMUObject *dsi_obj, +uint32_t min,uint32_t max,uint32_t *pasid) +{ +VFIOContainer *container =container_of (dsi_obj, VFIOContainer, dsi_obj); +struct vfio_iommu_type1_pasid_request req; +unsigned long argsz;
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1197
Quoted 10 lines
hw/vfio/common.c +10
-0+ + argsz =sizeof (req); + req.argsz = argsz; + req.flags =VFIO_IOMMU_PASID_ALLOC ; + req.alloc_pasid.min = min; + req.alloc_pasid.max = max; + +if (ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req)) { +error_report ("%s: %d, alloc failed" , __func__, -errno); +return -errno;
Note that errno is prone to change by other syscalls. Better cache it right after the ioctl.
Quoted 10 lines
hw/vfio/common.c +10
-0+ } + *pasid = req.alloc_pasid.result; +return 0 ; +} + +static int vfio_ds_iommu_pasid_free (DualStageIOMMUObject *dsi_obj, +uint32_t pasid) +{ +VFIOContainer *container =container_of (dsi_obj, VFIOContainer, dsi_obj); +struct vfio_iommu_type1_pasid_request req;
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1217
Quoted 10 lines
hw/vfio/common.c +10
-0+unsigned long argsz; + + argsz =sizeof (req); + req.argsz = argsz; + req.flags =VFIO_IOMMU_PASID_FREE ; + req.free_pasid = pasid; + +if (ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req)) { +error_report ("%s: %d, free failed" , __func__, -errno); +return -errno;
Same here.
Quoted 13 lines
hw/vfio/common.c +6
-0+ } +return 0 ; +} +static struct DualStageIOMMUOps vfio_ds_iommu_ops = { + .pasid_alloc = vfio_ds_iommu_pasid_alloc, + .pasid_free = vfio_ds_iommu_pasid_free, }; static int vfio_get_iommu_info(VFIOContainer *container,Signature · 1 line
2.7.4
Signature · 1 line
Peter Xu
RE: [RFC v3 12/25] vfio/common: add pasid_alloc/free support
Quoted 67 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 3:32 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 12/25] vfio/common: add pasid_alloc/free supportQuoted 14 lines · Yi L
From: Liu Yi L <[email protected]>
This patch adds VFIO pasid alloc/free support to allow host intercept in PASID allocation for VM by adding VFIO implementation of DualStageIOMMUOps.pasid_alloc/free callbacks.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Eric Auger <[email protected]>Cc Yi Sun <[email protected]>Cc David Gibson <[email protected]>Cc Alex Williamson <[email protected]>Signed-off-by Liu Yi L <[email protected]>Quoted 2 lines
--- hw/vfio/common.c | 42++++++++++++++++++++++++++++++++++++++++++
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer
Quoted 7 lines
1 file changed, 42 insertions(+)hw/vfio/common.c index renamed from hw/vfio/common.c +0
-0a07824b..014f4e7 100644
hw/vfio/common.c +0
-0@@ -1179,7 +1179,49 @@ static int vfio_get_iommu_type(VFIOContainer*container,
Quoted 9 lines
hw/vfio/common.c +6
-0return -EINVAL ; } +static int vfio_ds_iommu_pasid_alloc (DualStageIOMMUObject *dsi_obj, +uint32_t min,uint32_t max,uint32_t *pasid) +{ +VFIOContainer *container =container_of (dsi_obj, VFIOContainer, dsi_obj); +struct vfio_iommu_type1_pasid_request req; +unsigned long argsz;hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1197
Quoted 10 lines
hw/vfio/common.c +10
-0+ + argsz =sizeof (req); + req.argsz = argsz; + req.flags =VFIO_IOMMU_PASID_ALLOC ; + req.alloc_pasid.min = min; + req.alloc_pasid.max = max; + +if (ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req)) { +error_report ("%s: %d, alloc failed" , __func__, -errno); +return -errno;Note that errno is prone to change by other syscalls. Better cache it right after the ioctl.
Quoted 13 lines
hw/vfio/common.c +13
-0+ } + *pasid = req.alloc_pasid.result; +return 0 ; +} + +static int vfio_ds_iommu_pasid_free (DualStageIOMMUObject *dsi_obj, +uint32_t pasid) { +VFIOContainer *container =container_of (dsi_obj, VFIOContainer, dsi_obj); +struct vfio_iommu_type1_pasid_request req; +unsigned long argsz; + + argsz =sizeof (req); + req.argsz = argsz;
Quoted 8 lines
hw/vfio/common.c static int vfio_get_iommu_type(VFIOContainer *container, · line 1217
Quoted 6 lines
hw/vfio/common.c +6
-0+ req.flags =VFIO_IOMMU_PASID_FREE ; + req.free_pasid = pasid; + +if (ioctl (container->fd,VFIO_IOMMU_PASID_REQUEST , &req)) { +error_report ("%s: %d, free failed" , __func__, -errno); +return -errno;Same here.
Got the two comments. Thanks,
Regards,
Yi Liu
RFC v3 13/25 intel_iommu: modify x-scalable-mode to be string option
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
Intel 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"]"
- "legacy": gives support for SL page table
- "modern": gives support for FL page table, pasid, virtual command
- if not configured, means no scalable mode support, if not proper configured, will throw error
--- hw/i386/intel_iommu.c | 27 +++++++++++++++++++++++++-- hw/i386/intel_iommu_internal.h | 3 +++ include/hw/i386/intel_iommu.h | 2 ++ 3 files changed, 30 insertions(+), 2 deletions(-)
hw/i386/intel_iommu.c +25 -2
@@ -3078,7 +3078,7 @@ 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(), };
@@ -3708,8 +3708,11 @@ 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_reset_caches (s);
@@ -3845,6 +3848,26 @@ static bool vtd_decide_config(IntelIOMMUState *s, Error **errp)return false; } +if (s->scalable_mode_str && + (strcmp (s->scalable_mode_str,"modern" ) && +strcmp (s->scalable_mode_str,"legacy" ))) { +error_setg (errp,"Invalid x-scalable-mode config" ); +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
@@ -190,8 +190,11 @@#define VTD_ECAP_PT (1ULL << 6)#define VTD_ECAP_MHMV (15ULL << 20)#define VTD_ECAP_SRS (1ULL << 31) +#define VTD_ECAP_PSS (19ULL << 35) +#define VTD_ECAP_PASID (1ULL << 40)#define VTD_ECAP_SMTS (1ULL << 43)#define VTD_ECAP_SLTS (1ULL << 46) +#define VTD_ECAP_FLTS (1ULL << 47)/* CAP_REG */ /* (offset >> 4) << 24 */
include/hw/i386/intel_iommu.h +2 -0
@@ -244,6 +244,8 @@ 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
Re: [RFC v3 13/25] intel_iommu: modify x-scalable-mode to be string option
Quoted 9 lines · Yi L
From: Liu Yi L <[email protected]>
Intel 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"]"
Maybe also "off" when someone wants to explicitly disable it?
Quoted 35 lines
- "legacy": gives support for SL page table
- "modern": gives support for FL page table, pasid, virtual command
- if not configured, means no scalable mode support, if not proper configured, will throw errorCc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>Signed-off-by Yi Sun <[email protected]>--- hw/i386/intel_iommu.c | 27 +++++++++++++++++++++++++-- hw/i386/intel_iommu_internal.h | 3 +++ include/hw/i386/intel_iommu.h | 2 ++ 3 files changed, 30 insertions(+), 2 deletions(-)hw/i386/intel_iommu.c +1
-1@@ -3078,7 +3078,7 @@ 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(), };
hw/i386/intel_iommu.c static void vtd_init(IntelIOMMUState *s) · line 3715
Quoted 9 lines
hw/i386/intel_iommu.c +4
-1@@ -3708,8 +3708,11 @@ 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 ;
This patch might be good to be the last one after all the impls are ready.
hw/i386/intel_iommu.c static bool vtd_decide_config(IntelIOMMUState *s, Error **errp) · line 3854
Quoted 11 lines
hw/i386/intel_iommu.c +4
-0}vtd_reset_caches (s);@@ -3845,6 +3848,26 @@ static bool vtd_decide_config(IntelIOMMUState *s, Error **errp)return false; } +if (s->scalable_mode_str && + (strcmp (s->scalable_mode_str,"modern" ) && +strcmp (s->scalable_mode_str,"legacy" ))) { +error_setg (errp,"Invalid x-scalable-mode config" );
Maybe "..., Please use 'modern', 'legacy', or 'off'." to show options.
Quoted 51 lines
hw/i386/intel_iommu.c +16
-0+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@@ -190,8 +190,11 @@#define VTD_ECAP_PT (1ULL << 6)#define VTD_ECAP_MHMV (15ULL << 20)#define VTD_ECAP_SRS (1ULL << 31) +#define VTD_ECAP_PSS (19ULL << 35) +#define VTD_ECAP_PASID (1ULL << 40)#define VTD_ECAP_SMTS (1ULL << 43)#define VTD_ECAP_SLTS (1ULL << 46) +#define VTD_ECAP_FLTS (1ULL << 47)/* CAP_REG */ /* (offset >> 4) << 24 */ include/hw/i386/intel_iommu.h +2
-0@@ -244,6 +244,8 @@ 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
Signature · 1 line
Peter Xu
RE: [RFC v3 13/25] intel_iommu: modify x-scalable-mode to be string option
Quoted 11 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 3:44 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 13/25]intel_iommu: modify x-scalable-mode to be string optionQuoted 4 lines · Yi L
From: Liu Yi L <[email protected]>
Intel VT-d 3.0 introduces scalable mode, and it has a bunch of capabilities related to scalable mode translation, thus there are multiple
Quoted 8 lines
combinations.
Quoted 5 lines
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"]"
Maybe also "off" when someone wants to explicitly disable it?
emmm, I think x-scalable-mode should be disabled by default. It is enabled only when "legacy" or "modern" is configured. I'm fine to add "off" as an explicit way to turn it off if you think it is necessary. :-)
Quoted 40 lines
Quoted 26 lines
- "legacy": gives support for SL page table
- "modern": gives support for FL page table, pasid, virtual command
- if not configured, means no scalable mode support, if not proper configured, will throw errorCc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>Signed-off-by Yi Sun <[email protected]>--- hw/i386/intel_iommu.c | 27 +++++++++++++++++++++++++-- hw/i386/intel_iommu_internal.h | 3 +++ include/hw/i386/intel_iommu.h | 2 ++ 3 files changed, 30 insertions(+), 2 deletions(-)hw/i386/intel_iommu.c index renamed from hw/i386/intel_iommu.c +0
-01c1eb7f..33be40c 100644
hw/i386/intel_iommu.c static Property vtd_properties[] = { · line 3080
Quoted 3 lines
hw/i386/intel_iommu.c +0
-0@@ -3078,7 +3078,7 @@ 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),
hw/i386/intel_iommu.c static Property vtd_properties[] = { · line 3081
Quoted 1 line
hw/i386/intel_iommu.c +0
-1- DEFINE_PROP_BOOL("x-scalable-mode", IntelIOMMUState, scalable_mode,FALSE),
Quoted 8 lines
hw/i386/intel_iommu.c +2
-0+DEFINE_PROP_STRING ("x-scalable-mode" , IntelIOMMUState, + scalable_mode_str),DEFINE_PROP_BOOL ("dma-drain" , IntelIOMMUState, dma_drain, true), DEFINE_PROP_END_OF_LIST(), };@@ -3708,8 +3708,11 @@ static void vtd_init(IntelIOMMUState *s) }
Quoted 9 lines
hw/i386/intel_iommu.c static void vtd_init(IntelIOMMUState *s) · line 3715
Quoted 7 lines
hw/i386/intel_iommu.c +4
-1/* 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 ;This patch might be good to be the last one after all the impls are ready.
Oh, yes. Let me reorder it in next version.
Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_decide_config(IntelIOMMUState *s,
Quoted 4 lines
hw/i386/intel_iommu.c +0
-0}vtd_reset_caches (s);@@ -3845,6 +3848,26 @@ static bool vtd_decide_config(IntelIOMMUState *s,Error **errp)
Quoted 3 lines
hw/i386/intel_iommu.c +0
-0return false; }
Quoted 6 lines
hw/i386/intel_iommu.c static bool vtd_decide_config(IntelIOMMUState *s, Error **errp) · line 3854
Quoted 4 lines
hw/i386/intel_iommu.c +4
-0+if (s->scalable_mode_str && + (strcmp (s->scalable_mode_str,"modern" ) && +strcmp (s->scalable_mode_str,"legacy" ))) { +error_setg (errp,"Invalid x-scalable-mode config" );Maybe "..., Please use 'modern', 'legacy', or 'off'." to show options.
Got it.
Thanks,
Yi Liu
Re: [RFC v3 13/25] intel_iommu: modify x-scalable-mode to be string option
Quoted 16 lines · Yi L
Quoted 16 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 3:44 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 13/25]intel_iommu: modify x-scalable-mode to be string optionQuoted 4 lines · Yi L
From: Liu Yi L <[email protected]>
Intel VT-d 3.0 introduces scalable mode, and it has a bunch of capabilities related to scalable mode translation, thus there are multiple
combinations.
Quoted 4 lines
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:
Quoted 7 lines
Quoted 3 lines
Quoted 1 line
"-device intel-iommu,x-scalable-mode=["legacy"|"modern"]"
Maybe also "off" when someone wants to explicitly disable it?
emmm, I think x-scalable-mode should be disabled by default. It is enabled only when "legacy" or "modern" is configured. I'm fine to add "off" as an explicit way to turn it off if you think it is necessary. :-)
It's not necessary. It'll be necessary when we remove "x-" and change the default value. However it'll always be good to provide all options explicitly in the parameter starting from when we design it, imho. It's still experimental, so... Your call. :)
Signature · 1 line
Peter Xu
RE: [RFC v3 13/25] intel_iommu: modify x-scalable-mode to be string option
Quoted 33 lines
From: Peter Xu <[email protected]>
Sent: Thursday, February 13, 2020 12:06 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 13/25]intel_iommu: modify x-scalable-mode to be string optionQuoted 17 lines · Yi L
Quoted 17 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 3:44 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 13/25]intel_iommu: modify x-scalable-mode to be string optionQuoted 5 lines · Yi L
From: Liu Yi L <[email protected]>
Intel VT-d 3.0 introduces scalable mode, and it has a bunch of capabilities related to scalable mode translation, thus there are multiple
combinations.
Quoted 4 lines
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:
Quoted 8 lines
Quoted 3 lines
Quoted 1 line
"-device intel-iommu,x-scalable-mode=["legacy"|"modern"]"
Maybe also "off" when someone wants to explicitly disable it?
emmm, I think x-scalable-mode should be disabled by default. It is enabled only when "legacy" or "modern" is configured. I'm fine to add "off" as an explicit way to turn it off if you think it is necessary. :-)
Quoted 4 lines
It's not necessary. It'll be necessary when we remove "x-" and change the default value. However it'll always be good to provide all options explicitly in the parameter starting from when we design it, imho. It's still experimental, so... Your call. :)
Got it. Let me add it in next version. 😊
Regards,
Yi Liu
RFC v3 14/25 intel_iommu: add virtual command capability support
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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 | 163 ++++++++++++++++++++++++++++++++++++++++- hw/i386/intel_iommu_internal.h | 38 ++++++++++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 6 +- 4 files changed, 206 insertions(+), 2 deletions(-)
hw/i386/intel_iommu.c +162 -1
@@ -2649,6 +2649,142 @@ static void vtd_handle_iectl_write(IntelIOMMUState *s) } } +static int vtd_request_pasid_alloc (IntelIOMMUState *s,uint32_t *pasid) +{ +VTDBus *vtd_bus; +int bus_n, devfn, ret = -errno; +VTDIOMMUContext *vtd_icx; + +for (bus_n =0 ; bus_n <PCI_BUS_MAX ; bus_n++) { + vtd_bus =vtd_find_as_from_bus_num (s, bus_n); +if (!vtd_bus) { +continue ; + } +for (devfn =0 ; devfn <PCI_DEVFN_MAX ; devfn++) { + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +continue ; + } + +/* +* 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 object to do the same thing just +* to make sure the allocation happens only once. +*/ + ret =ds_iommu_pasid_alloc (vtd_icx->dsi_obj, +VTD_MIN_HPASID ,VTD_MAX_HPASID , pasid); +if (!ret) { +break ; + } + } + } +return ret; +} + +static int vtd_request_pasid_free (IntelIOMMUState *s,uint32_t pasid) +{ +VTDBus *vtd_bus; +int bus_n, devfn, ret = -errno; +VTDIOMMUContext *vtd_icx; + +for (bus_n =0 ; bus_n <PCI_BUS_MAX ; bus_n++) { + vtd_bus =vtd_find_as_from_bus_num (s, bus_n); +if (!vtd_bus) { +continue ; + } +for (devfn =0 ; devfn <PCI_DEVFN_MAX ; devfn++) { + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +continue ; + } +/* +* 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 +* object to do the same thing just to make sure the +* free happens only once. +*/ + ret =ds_iommu_pasid_free (vtd_icx->dsi_obj, pasid); +if (!ret) { +break ; + } + } + } +return ret; +} + +/* +* If IP is not set, set it and return 0 +* If IP is already set, return -1 +*/ +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;
@@ -2938,6 +3074,23 @@ 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);
@@ -3712,7 +3865,8 @@ static void vtd_init(IntelIOMMUState *s) 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_ECAP_FLTS |VTD_ECAP_PSS |VTD_ECAP_VCS ; + s->vccap |=VTD_VCCAP_PAS ; }vtd_reset_caches (s);
@@ -3768,6 +3922,13 @@ 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
hw/i386/intel_iommu_internal.h +38 -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 */
@@ -193,6 +199,7 @@#define VTD_ECAP_PSS (19ULL << 35)#define VTD_ECAP_PASID (1ULL << 40)#define VTD_ECAP_SMTS (1ULL << 43) +#define VTD_ECAP_VCS (1ULL << 44)#define VTD_ECAP_SLTS (1ULL << 46)#define VTD_ECAP_FLTS (1ULL << 47)
@@ -315,6 +322,37 @@ 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_MIN_HPASID 1 +#define VTD_MAX_HPASID 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 +5 -1
@@ -46,7 +46,7 @@#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
@@ -285,6 +285,10 @@ 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 */ +/* * Protects IOMMU states in general. Currently it protects the * per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace.
Signature · 1 line
2.7.4
Re: [RFC v3 14/25] intel_iommu: add virtual command capability support
Quoted 47 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>Signed-off-by Yi Sun <[email protected]>--- hw/i386/intel_iommu.c | 163 ++++++++++++++++++++++++++++++++++++++++- hw/i386/intel_iommu_internal.h | 38 ++++++++++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 6 +- 4 files changed, 206 insertions(+), 2 deletions(-)hw/i386/intel_iommu.c +17
-0@@ -2649,6 +2649,142 @@ static void vtd_handle_iectl_write(IntelIOMMUState *s) } } +static int vtd_request_pasid_alloc (IntelIOMMUState *s,uint32_t *pasid) +{ +VTDBus *vtd_bus; +int bus_n, devfn, ret = -errno; +VTDIOMMUContext *vtd_icx; + +for (bus_n =0 ; bus_n <PCI_BUS_MAX ; bus_n++) { + vtd_bus =vtd_find_as_from_bus_num (s, bus_n); +if (!vtd_bus) { +continue ; + } +for (devfn =0 ; devfn <PCI_DEVFN_MAX ; devfn++) { + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +continue ; + } +
hw/i386/intel_iommu.c static void vtd_handle_iectl_write(IntelIOMMUState *s) · line 2678
Quoted 10 lines
hw/i386/intel_iommu.c +10
-0+/* +* 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 object to do the same thing just +* to make sure the allocation happens only once. +*/ + ret =ds_iommu_pasid_alloc (vtd_icx->dsi_obj, +VTD_MIN_HPASID ,VTD_MAX_HPASID , pasid);
Your indents are always strange to me for long funcalls... Not a complaint though, as long as no one else complains. :)
Quoted 35 lines
hw/i386/intel_iommu.c +35
-0+if (!ret) { +break ; + } + } + } +return ret; +} + +static int vtd_request_pasid_free (IntelIOMMUState *s,uint32_t pasid) +{ +VTDBus *vtd_bus; +int bus_n, devfn, ret = -errno; +VTDIOMMUContext *vtd_icx; + +for (bus_n =0 ; bus_n <PCI_BUS_MAX ; bus_n++) { + vtd_bus =vtd_find_as_from_bus_num (s, bus_n); +if (!vtd_bus) { +continue ; + } +for (devfn =0 ; devfn <PCI_DEVFN_MAX ; devfn++) { + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +continue ; + } +/* +* 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 +* object to do the same thing just to make sure the +* free happens only once. +*/ + ret =ds_iommu_pasid_free (vtd_icx->dsi_obj, pasid); +if (!ret) {
hw/i386/intel_iommu.c static void vtd_handle_iectl_write(IntelIOMMUState *s) · line 2723
Quoted 10 lines
hw/i386/intel_iommu.c +10
-0+break ; + } + } + } +return ret; +} + +/* +* If IP is not set, set it and return 0 +* If IP is already set, return -1
Out of date? Instead can mention that this also resets the reply status code to zero implicitly so by default it will return a success.
Other than that:
Quoted 217 lines
hw/i386/intel_iommu.c +90
-1+*/ +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;@@ -2938,6 +3074,23 @@ 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);@@ -3712,7 +3865,8 @@ static void vtd_init(IntelIOMMUState *s) 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_ECAP_FLTS |VTD_ECAP_PSS |VTD_ECAP_VCS ; + s->vccap |=VTD_VCCAP_PAS ; }vtd_reset_caches (s);@@ -3768,6 +3922,13 @@ 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 hw/i386/intel_iommu_internal.h +38
-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 */ @@ -193,6 +199,7 @@#define VTD_ECAP_PSS (19ULL << 35)#define VTD_ECAP_PASID (1ULL << 40)#define VTD_ECAP_SMTS (1ULL << 43) +#define VTD_ECAP_VCS (1ULL << 44)#define VTD_ECAP_SLTS (1ULL << 46)#define VTD_ECAP_FLTS (1ULL << 47)@@ -315,6 +322,37 @@ 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_MIN_HPASID 1 +#define VTD_MAX_HPASID 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 +5
-1@@ -46,7 +46,7 @@#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@@ -285,6 +285,10 @@ 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 */ +/* * Protects IOMMU states in general. Currently it protects the * per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace. Signature · 1 line
2.7.4
Signature · 1 line
Peter Xu
RE: [RFC v3 14/25] intel_iommu: add virtual command capability support
Quoted 60 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 4:16 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 14/25]intel_iommu: add virtual command capability supportQuoted 15 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>Signed-off-by Yi Sun <[email protected]>Quoted 2 lines
--- hw/i386/intel_iommu.c | 163++++++++++++++++++++++++++++++++++++++++-
hw/i386/intel_iommu.c static void
Quoted 10 lines
hw/i386/intel_iommu_internal.h | 38 ++++++++++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 6 +- 4 files changed, 206 insertions(+), 2 deletions(-)hw/i386/intel_iommu.c index renamed from hw/i386/intel_iommu.c +0
-033be40c..43a728f 100644
hw/i386/intel_iommu.c +0
-0@@ -2649,6 +2649,142 @@ static void
vtd_handle_iectl_write(IntelIOMMUState *s)Quoted 25 lines
hw/i386/intel_iommu.c +22
-0} } +static int vtd_request_pasid_alloc (IntelIOMMUState *s,uint32_t +*pasid) { +VTDBus *vtd_bus; +int bus_n, devfn, ret = -errno; +VTDIOMMUContext *vtd_icx; + +for (bus_n =0 ; bus_n <PCI_BUS_MAX ; bus_n++) { + vtd_bus =vtd_find_as_from_bus_num (s, bus_n); +if (!vtd_bus) { +continue ; + } +for (devfn =0 ; devfn <PCI_DEVFN_MAX ; devfn++) { + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +continue ; + } + +/* +* 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
Quoted 8 lines
hw/i386/intel_iommu.c static void vtd_handle_iectl_write(IntelIOMMUState *s) · line 2678
Quoted 5 lines
hw/i386/intel_iommu.c +5
-0+* per-device iommu object to do the same thing just +* to make sure the allocation happens only once. +*/ + ret =ds_iommu_pasid_alloc (vtd_icx->dsi_obj, +VTD_MIN_HPASID ,VTD_MAX_HPASID , pasid);Your indents are always strange to me for long funcalls... Not a complaint though, as long as no one else complains. :)
yeah, I'm also not feeling well with them... I'll try to make the indents for long funccalls better. 😊
Quoted 41 lines
Quoted 41 lines
hw/i386/intel_iommu.c +40
-0+if (!ret) { +break ; + } + } + } +return ret; +} + +static int vtd_request_pasid_free (IntelIOMMUState *s,uint32_t pasid) +{ +VTDBus *vtd_bus; +int bus_n, devfn, ret = -errno; +VTDIOMMUContext *vtd_icx; + +for (bus_n =0 ; bus_n <PCI_BUS_MAX ; bus_n++) { + vtd_bus =vtd_find_as_from_bus_num (s, bus_n); +if (!vtd_bus) { +continue ; + } +for (devfn =0 ; devfn <PCI_DEVFN_MAX ; devfn++) { + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +continue ; + } +/* +* 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 +* object to do the same thing just to make sure the +* free happens only once. +*/ + ret =ds_iommu_pasid_free (vtd_icx->dsi_obj, pasid); +if (!ret) { +break ; + } + } + } +return ret;
Quoted 8 lines
hw/i386/intel_iommu.c static void vtd_handle_iectl_write(IntelIOMMUState *s) · line 2723
Quoted 5 lines
hw/i386/intel_iommu.c +5
-0+} + +/* +* If IP is not set, set it and return 0 +* If IP is already set, return -1 Out of date? Instead can mention that this also resets the reply status code to zero implicitly so by default it will return a success.
Ooops, yeah, it's out of date. Will fix it.
Quoted 4 lines
Other than that:
Reviewed-by Peter Xu <[email protected]>
Thanks a lot for the patient reviewing.
Regards,
Yi Liu
Quoted 223 lines
hw/i386/intel_iommu.c static void vtd_mem_write(void *opaque, hwaddr
Quoted 63 lines
hw/i386/intel_iommu.c +59
-0+*/ +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;@@ -2938,6 +3074,23 @@ static void vtd_mem_write(void *opaque, hwaddraddr,
Quoted 52 lines
hw/i386/intel_iommu.c +17
-0vtd_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); @@ -3712,7 +3865,8 @@ staticvoid
vtd_init(IntelIOMMUState *s)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_PASIDhw/i386/intel_iommu.c +9
-1- | VTD_ECAP_FLTS | VTD_ECAP_PSS; + | VTD_ECAP_FLTS | VTD_ECAP_PSS | VTD_ECAP_VCS; + s->vccap |=VTD_VCCAP_PAS ; }vtd_reset_caches (s);@@ -3768,6 +3922,13 @@ 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 diff --git
a/hw/i386/intel_iommu_internal.hb/hw/i386/intel_iommu_internal.hindex c4dbb2c..fb5fdc2 100644hw/i386/intel_iommu_internal.h line 88
Quoted 4 lines
hw/i386/intel_iommu_internal.h +1
-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 */
hw/i386/intel_iommu_internal.h
Quoted 4 lines
hw/i386/intel_iommu_internal.h +4
-0+#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 */
hw/i386/trace-events vtd_reg_write_gcmd(uint32_t status, uint32_t val) "status
Quoted 56 lines
hw/i386/intel_iommu_internal.h +34
-0+#define DMAR_VCRSP_REG_HI 0xE24/* IOTLB registers */ #define DMAR_IOTLB_REG_OFFSET 0xf0/* Offset to the IOTLB registers */ @@ -193,6 +199,7 @@#define VTD_ECAP_PSS (19ULL << 35)#define VTD_ECAP_PASID (1ULL << 40)#define VTD_ECAP_SMTS (1ULL << 43) +#define VTD_ECAP_VCS (1ULL << 44)#define VTD_ECAP_SLTS (1ULL << 46)#define VTD_ECAP_FLTS (1ULL << 47)@@ -315,6 +322,37 @@ 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_MIN_HPASID 1 +#define VTD_MAX_HPASID 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 index renamed from hw/i386/trace-events +0
-0e48bef2..71536a7 100644
hw/i386/trace-events +0
-0@@ -51,6 +51,7 @@ vtd_reg_write_gcmd(uint32_t status, uint32_t val) "status0x%"PRIx32" value 0x%"
Quoted 7 lines
hw/i386/trace-events +0
-0vtd_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)""hw/i386/trace-events +2
-0+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_tiova,
uint64_tgpa,uint64_tmask) "dev %02x:%02x.%02x iova 0x%"PRIx64" ->gpa 0x%"PRIx64" mask 0x%"PRIx64
vtd_dmar_enable(bool en)"enable %d"Quoted 1 line
vtd_dmar_fault(uint16_t sid, int fault, uint64_t addr, bool is_write)"sid0x%"PRIx16" fault %d addr 0x%"PRIx64" write %d"
Quoted 25 lines
(quoted diff) +0
-0
b/include/hw/i386/intel_iommu.hindex 1ef2917..4158116 100644include/hw/i386/intel_iommu.h +5
-2@@ -46,7 +46,7 @@#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@@ -285,6 +285,10 @@ 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 */ +/* * Protects IOMMU states in general. Currently it protects the * per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace. -- Quoted 2 lines
2.7.4
--
Peter Xu
Re: [RFC v3 14/25] intel_iommu: add virtual command capability support
In reply to Liu, Yi Lhw/i386/intel_iommu_internal.h typedef enum VTDFaultReason { · line 336
Quoted 9 lines · Yi L
hw/i386/intel_iommu_internal.h +9
-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_MIN_HPASID 1 +#define VTD_MAX_HPASID 0xFFFFF
One more question: I see that PASID is defined as 20bits long. It's fine. However I start to get confused on how the Scalable Mode PASID Directory could service that much of PASID entries.
I'm looking at spec 3.4.3, Figure 3-8.
Firstly, we only have two levels for a PASID table. The context entry of a device stores a pointer to the "Scalable Mode PASID Directory" page. I see that there're 2^14 entries in "Scalable Mode PASID Directory" page, each is a "Scalable Mode PASID Table". However... how do we fit in the 4K page if each entry is a pointer of x86_64 (8 bytes) while there're 2^14 entries? A simple math gives me 4K/8 = 512, which means the "Scalable Mode PASID Directory" page can only have 512 entries, then how the 2^14 come from? Hmm??
Apart of this: also I just noticed (when reading the latter part of the series) that the time that a pasid table walk can consume will depend on this value too. I'd suggest to make this as small as we can, as long as it satisfies the usage. We can even bump it in the future.
Signature · 1 line
Peter Xu
RE: [RFC v3 14/25] intel_iommu: add virtual command capability support
Quoted 22 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 5:57 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 14/25]intel_iommu: add virtual command capability supporthw/i386/intel_iommu_internal.h typedef enum VTDFaultReason { · line 336
Quoted 9 lines · Yi L
hw/i386/intel_iommu_internal.h +9
-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_MIN_HPASID 1 +#define VTD_MAX_HPASID 0xFFFFFOne more question: I see that PASID is defined as 20bits long. It's fine. However I start to get confused on how the Scalable Mode PASID Directory could service that much of PASID entries.
I'm looking at spec 3.4.3, Figure 3-8.
Quoted 8 lines
Firstly, we only have two levels for a PASID table. The context entry of a device stores a pointer to the "Scalable Mode PASID Directory" page. I see that there're 2^14 entries in "Scalable Mode PASID Directory" page, each is a "Scalable Mode PASID Table". However... how do we fit in the 4K page if each entry is a pointer of
x86_64(8 bytes) while there're 2^14 entries? A simple math gives me 4K/8 = 512, which means the "Scalable Mode PASID Directory" page can only have 512 entries, then how the 2^14 come from? Hmm??
I checked with Kevin. The spec doesn't say the dir table is 4K. It says 4K only for pasid table. Also, if you look at 9.4, scalabe-mode context entry includes a PDTS field to specify the actual size of the directory table.
Quoted 5 lines
Apart of this: also I just noticed (when reading the latter part of the series) that the time that a pasid table walk can consume will depend on this value too. I'd suggest to make this as small as we can, as long as it satisfies the usage. We can even bump it in the future.
I see. This looks to be an optimization. right? Instead of modify the value of this macro, I think we can do this optimization by tracking the allocated PASIDs in QEMU. Thus, the pasid table walk would be more efficient and also no dependency on the VTD_MAX_HPASID. Does it make sense to you? :-)
Regards,
Yi Liu
Re: [RFC v3 14/25] intel_iommu: add virtual command capability support
Quoted 31 lines · Yi L
Quoted 22 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 5:57 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 14/25]intel_iommu: add virtual command capability supporthw/i386/intel_iommu_internal.h typedef enum VTDFaultReason { · line 336
Quoted 9 lines · Yi L
hw/i386/intel_iommu_internal.h +9
-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_MIN_HPASID 1 +#define VTD_MAX_HPASID 0xFFFFFOne more question: I see that PASID is defined as 20bits long. It's fine. However I start to get confused on how the Scalable Mode PASID Directory could service that much of PASID entries.
I'm looking at spec 3.4.3, Figure 3-8.
Quoted 8 lines
Firstly, we only have two levels for a PASID table. The context entry of a device stores a pointer to the "Scalable Mode PASID Directory" page. I see that there're 2^14 entries in "Scalable Mode PASID Directory" page, each is a "Scalable Mode PASID Table". However... how do we fit in the 4K page if each entry is a pointer of
x86_64(8 bytes) while there're 2^14 entries? A simple math gives me 4K/8 = 512, which means the "Scalable Mode PASID Directory" page can only have 512 entries, then how the 2^14 come from? Hmm??
Quoted 3 lines
I checked with Kevin. The spec doesn't say the dir table is 4K. It says 4K only for pasid table. Also, if you look at 9.4, scalabe-mode context entry includes a PDTS field to specify the actual size of the directory table.
Ah I see. Then it seems to be lost then in this series. Say, I think vtd_sm_pasid_table_walk() should also stop walking until reaching the size there, and you need to fetch that size info from the context entry before walk starts.
Quoted 7 lines
Quoted 6 lines
Apart of this: also I just noticed (when reading the latter part of the series) that the time that a pasid table walk can consume will depend on this value too. I'd suggest to make this as small as we can, as long as it satisfies the usage. We can even bump it in the future.
Quoted 5 lines
I see. This looks to be an optimization. right? Instead of modify the value of this macro, I think we can do this optimization by tracking the allocated PASIDs in QEMU. Thus, the pasid table walk would be more efficient and also no dependency on the
VTD_MAX_HPASID. Does it make sense to you? :-)
Yeah sounds good. :)
Thanks,
Signature · 1 line
Peter Xu
Re: [RFC v3 14/25] intel_iommu: add virtual command capability support
On Thu, Feb 13, 2020 at 09:31:10AM -0500, Peter Xu wrote:
[...]
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Apart of this: also I just noticed (when reading the latter part of the series) that the time that a pasid table walk can consume will depend on this value too. I'd suggest to make this as small as we can, as long as it satisfies the usage. We can even bump it in the future.
Quoted 7 lines
Quoted 5 lines
I see. This looks to be an optimization. right? Instead of modify the value of this macro, I think we can do this optimization by tracking the allocated PASIDs in QEMU. Thus, the pasid table walk would be more efficient and also no dependency on the
VTD_MAX_HPASID. Does it make sense to you? :-)Yeah sounds good. :)
Just to make sure it's safe even for when the global allocation is not happening (full emulation devices? Do they need the PASID table walk too?). Anyway, be careful to not miss some valid PASID entries, or we can still use the MIN(PASID_MAX, CONTEXT_ENTRY_SIZE) to be safe as a first version. Thanks,
Signature · 1 line
Peter Xu
RE: [RFC v3 14/25] intel_iommu: add virtual command capability support
Quoted 21 lines
From: Peter Xu < [email protected] >
Sent: Thursday, February 13, 2020 11:09 PM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 14/25]intel_iommu: add virtual command capability supportOn Thu, Feb 13, 2020 at 09:31:10AM -0500, Peter Xu wrote:
[...]
Quoted 12 lines
Quoted 11 lines
Quoted 5 lines
Apart of this: also I just noticed (when reading the latter part of the series) that the time that a pasid table walk can consume will depend on this value too. I'd suggest to make this as small as we can, as long as it satisfies the usage. We can even bump it in the future.
I see. This looks to be an optimization. right? Instead of modify the value of this macro, I think we can do this optimization by tracking the allocated PASIDs in QEMU. Thus, the pasid table walk would be more efficient and also no dependency on the
VTD_MAX_HPASID. Does it make sense to you? :-)
Quoted 5 lines
Quoted 1 line
Yeah sounds good. :)
Just to make sure it's safe even for when the global allocation is not happening (full emulation devices? Do they need the PASID table walk too?).
I'd say no. For full emulation devices, just needs to ensure the pasid cache is latest (do what guest told). Even the invalidation flushes too much cache, it just affects the performance but no correctness issue. This is different with passthru devices, if unbind too much, it means some passthru devices may encounter DMA fault later.
Quoted 3 lines
Anyway, be careful to not miss some valid PASID entries, or we can still use the
MIN(PASID_MAX, CONTEXT_ENTRY_SIZE)to be safe as a first version. Thanks,
Agreed. First version to ensure 100% safe.
Regards,
Yi Liu
RFC v3 15/25 intel_iommu: process pasid cache invalidation
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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.
--- hw/i386/intel_iommu.c | 66 ++++++++++++++++++++++++++++++++++++++---- hw/i386/intel_iommu_internal.h | 12 ++++++++ hw/i386/trace-events | 3 ++ 3 files changed, 76 insertions(+), 5 deletions(-)
hw/i386/intel_iommu.c +61 -5
@@ -2393,6 +2393,63 @@ static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc)return true; } +static int vtd_pasid_cache_dsi (IntelIOMMUState *s,uint16_t domain_id) +{ +return 0 ; +} + +static int vtd_pasid_cache_psi (IntelIOMMUState *s, +uint16_t domain_id,uint32_t pasid) +{ +return 0 ; +} + +static int vtd_pasid_cache_gsi (IntelIOMMUState *s) +{ +return 0 ; +} + +static bool vtd_process_pasid_desc (IntelIOMMUState *s, +VTDInvDesc *inv_desc) +{ +uint16_t domain_id; +uint32_t pasid; +int ret =0 ; + +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; + } + + 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 : + ret =vtd_pasid_cache_dsi (s, domain_id); +break ; + +case VTD_INV_DESC_PASIDC_PASID_SI : + ret =vtd_pasid_cache_psi (s, domain_id, pasid); +break ; + +case VTD_INV_DESC_PASIDC_GLOBAL : + ret =vtd_pasid_cache_gsi (s); +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 (ret ==0 ) ? true : false; +} +static bool vtd_process_inv_iec_desc (IntelIOMMUState *s,VTDInvDesc *inv_desc) {
@@ -2499,12 +2556,11 @@ 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
@@ -448,6 +448,18 @@ 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
Re: [RFC v3 15/25] intel_iommu: process pasid cache invalidation
Quoted 8 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Quoted 8 lines
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>
Signature · 1 line
Peter Xu
RE: [RFC v3 15/25] intel_iommu: process pasid cache invalidation
Quoted 23 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 4:17 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 15/25]intel_iommu: process pasid cache invalidationQuoted 8 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Quoted 8 lines
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>
Quoted 1 line
Reviewed-by Peter Xu <[email protected]>
Thanks😊
Regards,
Yi Liu
RFC v3 16/25 intel_iommu: add PASID cache management infrastructure
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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.
--- hw/i386/intel_iommu.c | 367 +++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 14 ++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 36 +++- 4 files changed, 417 insertions(+), 1 deletion(-)
hw/i386/intel_iommu.c +367 -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 +690,11 @@ 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 bool vtd_pdire_present (VTDPASIDDirEntry *pdire) {return pdire->val &1 ;
@@ -2393,19 +2402,370 @@ 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 is used to clear pasid_cache_gen of 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; + +if (!(pc_entry->pasid_cache_gen == s->pasid_cache_gen)) { +return false; + } + +switch (pc_info->flags &VTD_PASID_CACHE_INFO_MASK ) { +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 : +break ; +default : +return false; + } + +/* +* 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; + } +/* Compare cached pasid entry and latest pasid entry */ +if (!vtd_pasid_entry_compare (&pe, &pc_entry->pasid_entry)) { +/* pasid entry was updated, thus update the pasid cache */ + pc_entry->pasid_entry = pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +/* +* 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 false; +remove: +/* +* TODO: +* - send pasid unbind to host for passthru devices +* - when pasid-base-iotlb(piotlb) infrastructure is ready, +* should invalidate QEMU piotlb togehter with this change. +*/ +return true; +} +static int vtd_pasid_cache_dsi (IntelIOMMUState *s,uint16_t domain_id) { +VTDPASIDCacheInfo pc_info; + +trace_vtd_pasid_cache_dsi (domain_id); + + pc_info.flags =VTD_PASID_CACHE_DOMSI ; + pc_info.domain_id = domain_id; + +/* +* Loop all existing pasid caches and update them. +*/ +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + +/* +* TODO: Domain selective PASID cache invalidation +* flushes all the pasid caches within a domain. To +* be safe, after invalidating the pasid caches, emulator +* needs to replay the pasid bindings by walking guest +* pasid dir and pasid table. +*/ +vtd_iommu_unlock (s);return 0 ; } +/** +* 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, +bool allocate) +{ +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 && allocate) { + 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->context_cache_entry.context_cache_gen =0 ; + vtd_pasid_as->pasid = pasid; + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen =0 ; +g_hash_table_insert (s->vtd_pasid_as, new_key, vtd_pasid_as); + } +return vtd_pasid_as; +} + +/** +* This function updates the pasid entry cached in &vtd_pasid_as. +* Caller of this function should hold iommu_lock. +*/ +static inline void vtd_fill_in_pe_cache ( +VTDPASIDAddressSpace *vtd_pasid_as,VTDPASIDEntry *pe) +{ +IntelIOMMUState *s = vtd_pasid_as->iommu_state; +VTDPASIDCacheEntry *pc_entry = &vtd_pasid_as->pasid_cache_entry; + + pc_entry->pasid_entry = *pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +} +static int vtd_pasid_cache_psi (IntelIOMMUState *s,uint16_t domain_id,uint32_t pasid) { +VTDPASIDCacheInfo pc_info; +VTDPASIDEntry pe; +VTDBus *vtd_bus; +int bus_n, devfn; +VTDPASIDAddressSpace *vtd_pasid_as; +VTDIOMMUContext *vtd_icx; + +/* PASID selective implies a DID selective */ + pc_info.flags =VTD_PASID_CACHE_PASIDSI ; + pc_info.domain_id = domain_id; + pc_info.pasid = pasid; + +/* +* Regards to a pasid selective pasid cache invalidation (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 +* +* Here the handling of a PSI is: +* 1) loop all the exisitng 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) +* However, it is not good to always loop all devices. In this +* implementation. We do it in this ways: +* - For devices which have VTDIOMMUContext 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 VTDIOMMUContext +* instances, it is not necessary to create pasid cache at +* this phase since it could be created when vIOMMU do DMA +* address translation. This is not implemented yet since +* no PASID-capable emulated devices today. If we have it +* in future, the pasid cache shall be created there. +*/ + +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + + QLIST_FOREACH(vtd_icx, &s->vtd_dev_icx_list, next) { + vtd_bus = vtd_icx->vtd_bus; + devfn = vtd_icx->devfn; + bus_n =pci_bus_num (vtd_bus->bus); + +/* Step 1: fetch vtd_pasid_as and check if it is valid */ + vtd_pasid_as =vtd_add_find_pasid_as (s, vtd_bus, + devfn, pasid, true); +if (vtd_pasid_as && + (s->pasid_cache_gen == + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen)) { +/* +* pasid_cache_gen equals to s->pasid_cache_gen means +* vtd_pasid_as is valid after the above s->vtd_pasid_as +* updates. Thus no need for the below steps. +*/ +continue ; + } + +/* +* Step 2: vtd_pasid_as is not valid, it's potentailly a +* new pasid bind. Fetch guest pasid entry. +*/ +if (vtd_dev_get_pe_from_pasid (s, bus_n, devfn, pasid, &pe)) { +continue ; + } + +/* +* Step 3: pasid entry exists, update pasid cache +* +* Here need to check domain ID since guest pasid entry +* exists. What needs to do are: +* - update the pc_entry in the vtd_pasid_as +* - set proper pc_entry.pasid_cache_gen +* - pass down the latest guest pasid entry config to host +* (will be added in later patch) +*/ +if (domain_id ==vtd_pe_get_domain_id (&pe)) { +vtd_fill_in_pe_cache (vtd_pasid_as, &pe); + } + } +vtd_iommu_unlock (s);return 0 ; } +/** +* 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.flags =VTD_PASID_CACHE_GLOBAL ; + +/* +* Reset pasid cache is a big hammer, so use +* g_hash_table_foreach_remove which will free +* the vtd_pasid_as instances. +*/ +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + s->pasid_cache_gen =1 ; +} +static int vtd_pasid_cache_gsi (IntelIOMMUState *s) { +trace_vtd_pasid_cache_gsi (); + +vtd_iommu_lock (s); +vtd_pasid_cache_reset (s); + +/* +* TODO: Global PASID cache invalidation may be +* flushes all the pasid caches. To be safe, after +* invalidating the pasid caches, emulator needs +* to replay the pasid bindings by walking guest +* pasid dir and pasid table. +*/ +vtd_iommu_unlock (s);return 0 ; }
@@ -3659,8 +4019,11 @@ static int vtd_icx_register_ds_iommu(IOMMUContext *iommu_ctx,VTDIOMMUContext *vtd_dev_icx =container_of (iommu_ctx, VTDIOMMUContext, iommu_context); +IntelIOMMUState *s = vtd_dev_icx->iommu_state; vtd_dev_icx->dsi_obj = dsi_obj; +QLIST_INSERT_HEAD (&s->vtd_dev_icx_list, vtd_dev_icx, next); +return 0 ; }
@@ -3672,6 +4035,7 @@ static void vtd_icx_unregister_ds_iommu(IOMMUContext *iommu_ctx, iommu_context); vtd_dev_icx->dsi_obj =NULL ; +QLIST_REMOVE (vtd_dev_icx, next); }IOMMUContextOps vtd_iommu_context_ops = {
@@ -4130,6 +4494,7 @@ 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,
@@ -4155,6 +4520,8 @@ 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 +14 -0
@@ -311,6 +311,7 @@ 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.
@@ -485,6 +486,19 @@ struct VTDRootEntry { };typedef struct VTDRootEntry VTDRootEntry ; +struct VTDPASIDCacheInfo { +#define VTD_PASID_CACHE_GLOBAL (1ULL << 0) +#define VTD_PASID_CACHE_DOMSI (1ULL << 1) +#define VTD_PASID_CACHE_PASIDSI (1ULL << 2) +uint32_t flags; +uint16_t domain_id; +uint32_t pasid; +}; +#define VTD_PASID_CACHE_INFO_MASK (VTD_PASID_CACHE_GLOBAL | \ + VTD_PASID_CACHE_DOMSI | \ + VTD_PASID_CACHE_PASIDSI) +typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ; +/* Masks for struct VTDRootEntry */ #define VTD_ROOT_ENTRY_P 1ULL#define VTD_ROOT_ENTRY_CTP (~0xfffULL)
hw/i386/trace-events +1 -0
@@ -22,6 +22,7 @@ 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_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
include/hw/i386/intel_iommu.h +35 -1
@@ -69,6 +69,8 @@ typedef union VTD_IR_MSIAddress VTD_IR_MSIAddress;typedef struct VTDPASIDDirEntry VTDPASIDDirEntry ;typedef struct VTDPASIDEntry VTDPASIDEntry ;typedef struct VTDIOMMUContext VTDIOMMUContext ; +typedef struct VTDPASIDCacheEntry VTDPASIDCacheEntry ; +typedef struct VTDPASIDAddressSpace VTDPASIDAddressSpace ;/* Context-Entry */ struct VTDContextEntry {
@@ -101,6 +103,31 @@ struct VTDPASIDEntry {uint64_t val[8 ]; }; +struct pasid_key { +uint32_t pasid; +uint16_t sid; +}; + +struct VTDPASIDCacheEntry { +/* +* The cache entry is obsolete if +* pasid_cache_gen!=IntelIOMMUState.pasid_cache_gen +*/ +uint32_t pasid_cache_gen; +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;
@@ -122,6 +149,7 @@ struct VTDIOMMUContext {uint8_t devfn;IOMMUContext iommu_context;DualStageIOMMUObject *dsi_obj; +QLIST_ENTRY (VTDIOMMUContext) next;IntelIOMMUState *iommu_state; };
@@ -272,9 +300,14 @@ 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 */ +uint32_t pasid_cache_gen;/* Should be in [1,MAX] */ /* list of registered notifiers */ QLIST_HEAD(, VTDAddressSpace) vtd_as_with_notifiers; +/* list of VTDIOMMUContexts with DualStageIOMMUObject registered */ + QLIST_HEAD(, VTDIOMMUContext) vtd_dev_icx_list; +/* interrupt remapping */ bool intr_enabled;/* Whether guest enabled IR */ dma_addr_t intr_root;/* Interrupt remapping table pointer */
@@ -291,7 +324,8 @@ struct IntelIOMMUState {/* * Protects IOMMU states in general. Currently it protects the -* per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace. +* per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace, +* and pasid cache in VTDPASIDAddressSpace. */ QemuMutex iommu_lock; };
Signature · 1 line
2.7.4
Re: [RFC v3 16/25] intel_iommu: add PASID cache management infrastructure
Quoted 190 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>--- hw/i386/intel_iommu.c | 367 +++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 14 ++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 36 +++- 4 files changed, 417 insertions(+), 1 deletion(-)hw/i386/intel_iommu.c +96
-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 +690,11 @@ 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 bool vtd_pdire_present (VTDPASIDDirEntry *pdire) {return pdire->val &1 ;@@ -2393,19 +2402,370 @@ 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 is used to clear pasid_cache_gen of 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; +
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2501
Quoted 10 lines
hw/i386/intel_iommu.c +10
-0+if (!(pc_entry->pasid_cache_gen == s->pasid_cache_gen)) { +return false; + } + +switch (pc_info->flags &VTD_PASID_CACHE_INFO_MASK ) { +case VTD_PASID_CACHE_PASIDSI : +if (pc_info->pasid != pasid) { +return false; + } +/* Fall through */
Why fall through?
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2506
Quoted 5 lines
hw/i386/intel_iommu.c +5
-0+case VTD_PASID_CACHE_DOMSI : +if (pc_info->domain_id != did) { +return false; + } +/* Fall through */
Same here.
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2509
Quoted 3 lines
hw/i386/intel_iommu.c +3
-0+case VTD_PASID_CACHE_GLOBAL : +break ; +default :
Nevee reach here right? If so we can abort.
Quoted 55 lines
hw/i386/intel_iommu.c +53
-0+return false; + } + +/* +* 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; + } +/* Compare cached pasid entry and latest pasid entry */ +if (!vtd_pasid_entry_compare (&pe, &pc_entry->pasid_entry)) { +/* pasid entry was updated, thus update the pasid cache */ + pc_entry->pasid_entry = pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +/* +* 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 false; +remove: +/* +* TODO: +* - send pasid unbind to host for passthru devices +* - when pasid-base-iotlb(piotlb) infrastructure is ready, +* should invalidate QEMU piotlb togehter with this change. +*/ +return true; +} +static int vtd_pasid_cache_dsi (IntelIOMMUState *s,uint16_t domain_id) { +VTDPASIDCacheInfo pc_info; + +trace_vtd_pasid_cache_dsi (domain_id); + + pc_info.flags =VTD_PASID_CACHE_DOMSI ; + pc_info.domain_id = domain_id; + +/* +* Loop all existing pasid caches and update them. +*/
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2574
Quoted 10 lines
hw/i386/intel_iommu.c +10
-0+vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + +/* +* TODO: Domain selective PASID cache invalidation +* flushes all the pasid caches within a domain. To +* be safe, after invalidating the pasid caches, emulator +* needs to replay the pasid bindings by walking guest +* pasid dir and pasid table.
Better spell out on what special case we're handling here: When the guest setup a new PASID entry then send a PASID DSI.
Quoted 111 lines
hw/i386/intel_iommu.c +105
-0+*/ +vtd_iommu_unlock (s);return 0 ; } +/** +* 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, +bool allocate) +{ +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 && allocate) { + 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->context_cache_entry.context_cache_gen =0 ; + vtd_pasid_as->pasid = pasid; + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen =0 ; +g_hash_table_insert (s->vtd_pasid_as, new_key, vtd_pasid_as); + } +return vtd_pasid_as; +} + +/** +* This function updates the pasid entry cached in &vtd_pasid_as. +* Caller of this function should hold iommu_lock. +*/ +static inline void vtd_fill_in_pe_cache ( +VTDPASIDAddressSpace *vtd_pasid_as,VTDPASIDEntry *pe) +{ +IntelIOMMUState *s = vtd_pasid_as->iommu_state; +VTDPASIDCacheEntry *pc_entry = &vtd_pasid_as->pasid_cache_entry; + + pc_entry->pasid_entry = *pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +} +static int vtd_pasid_cache_psi (IntelIOMMUState *s,uint16_t domain_id,uint32_t pasid) { +VTDPASIDCacheInfo pc_info; +VTDPASIDEntry pe; +VTDBus *vtd_bus; +int bus_n, devfn; +VTDPASIDAddressSpace *vtd_pasid_as; +VTDIOMMUContext *vtd_icx; + +/* PASID selective implies a DID selective */ + pc_info.flags =VTD_PASID_CACHE_PASIDSI ; + pc_info.domain_id = domain_id; + pc_info.pasid = pasid; + +/* +* Regards to a pasid selective pasid cache invalidation (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 +* +* Here the handling of a PSI is: +* 1) loop all the exisitng 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) +* However, it is not good to always loop all devices. In this +* implementation. We do it in this ways: +* - For devices which have VTDIOMMUContext 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 VTDIOMMUContext +* instances, it is not necessary to create pasid cache at +* this phase since it could be created when vIOMMU do DMA +* address translation. This is not implemented yet since +* no PASID-capable emulated devices today. If we have it +* in future, the pasid cache shall be created there. +*/ + +vtd_iommu_lock (s); + g_hash_table_foreach_remove(s->vtd_pasid_as,
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2695
Quoted 10 lines
hw/i386/intel_iommu.c +10
-0+ vtd_flush_pasid, &pc_info); + + QLIST_FOREACH(vtd_icx, &s->vtd_dev_icx_list, next) { + vtd_bus = vtd_icx->vtd_bus; + devfn = vtd_icx->devfn; + bus_n =pci_bus_num (vtd_bus->bus); + +/* Step 1: fetch vtd_pasid_as and check if it is valid */ + vtd_pasid_as =vtd_add_find_pasid_as (s, vtd_bus, + devfn, pasid, true);
I feel like you wanted to pass "false" here for "allocate".
Quoted 54 lines
hw/i386/intel_iommu.c +51
-0+if (vtd_pasid_as && + (s->pasid_cache_gen == + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen)) { +/* +* pasid_cache_gen equals to s->pasid_cache_gen means +* vtd_pasid_as is valid after the above s->vtd_pasid_as +* updates. Thus no need for the below steps. +*/ +continue ; + } + +/* +* Step 2: vtd_pasid_as is not valid, it's potentailly a +* new pasid bind. Fetch guest pasid entry. +*/ +if (vtd_dev_get_pe_from_pasid (s, bus_n, devfn, pasid, &pe)) { +continue ; + } + +/* +* Step 3: pasid entry exists, update pasid cache +* +* Here need to check domain ID since guest pasid entry +* exists. What needs to do are: +* - update the pc_entry in the vtd_pasid_as +* - set proper pc_entry.pasid_cache_gen +* - pass down the latest guest pasid entry config to host +* (will be added in later patch) +*/ +if (domain_id ==vtd_pe_get_domain_id (&pe)) { +vtd_fill_in_pe_cache (vtd_pasid_as, &pe); + } + } +vtd_iommu_unlock (s);return 0 ; } +/** +* 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.flags =VTD_PASID_CACHE_GLOBAL ; + +/* +* Reset pasid cache is a big hammer, so use +* g_hash_table_foreach_remove which will free +* the vtd_pasid_as instances. +*/ + g_hash_table_foreach_remove(s->vtd_pasid_as,
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2759
Quoted 10 lines
hw/i386/intel_iommu.c +8
-0+ vtd_flush_pasid, &pc_info); + s->pasid_cache_gen =1 ; +} +static int vtd_pasid_cache_gsi (IntelIOMMUState *s) { +trace_vtd_pasid_cache_gsi (); + +vtd_iommu_lock (s); +vtd_pasid_cache_reset (s);
[1]
Quoted 123 lines
hw/i386/intel_iommu.c +16
-0+ +/* +* TODO: Global PASID cache invalidation may be +* flushes all the pasid caches. To be safe, after +* invalidating the pasid caches, emulator needs +* to replay the pasid bindings by walking guest +* pasid dir and pasid table. +*/ +vtd_iommu_unlock (s);return 0 ; }@@ -3659,8 +4019,11 @@ static int vtd_icx_register_ds_iommu(IOMMUContext *iommu_ctx,VTDIOMMUContext *vtd_dev_icx =container_of (iommu_ctx, VTDIOMMUContext, iommu_context); +IntelIOMMUState *s = vtd_dev_icx->iommu_state; vtd_dev_icx->dsi_obj = dsi_obj; +QLIST_INSERT_HEAD (&s->vtd_dev_icx_list, vtd_dev_icx, next); +return 0 ; }@@ -3672,6 +4035,7 @@ static void vtd_icx_unregister_ds_iommu(IOMMUContext *iommu_ctx, iommu_context); vtd_dev_icx->dsi_obj =NULL ; +QLIST_REMOVE (vtd_dev_icx, next); }IOMMUContextOps vtd_iommu_context_ops = {@@ -4130,6 +4494,7 @@ 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,@@ -4155,6 +4520,8 @@ 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 +14
-0@@ -311,6 +311,7 @@ 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. @@ -485,6 +486,19 @@ struct VTDRootEntry { };typedef struct VTDRootEntry VTDRootEntry ; +struct VTDPASIDCacheInfo { +#define VTD_PASID_CACHE_GLOBAL (1ULL << 0) +#define VTD_PASID_CACHE_DOMSI (1ULL << 1) +#define VTD_PASID_CACHE_PASIDSI (1ULL << 2) +uint32_t flags; +uint16_t domain_id; +uint32_t pasid; +}; +#define VTD_PASID_CACHE_INFO_MASK (VTD_PASID_CACHE_GLOBAL | \ + VTD_PASID_CACHE_DOMSI | \ + VTD_PASID_CACHE_PASIDSI) +typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ; +/* Masks for struct VTDRootEntry */ #define VTD_ROOT_ENTRY_P 1ULL#define VTD_ROOT_ENTRY_CTP (~0xfffULL)hw/i386/trace-events +1
-0@@ -22,6 +22,7 @@ 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_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%"PRIx32include/hw/i386/intel_iommu.h +15
-0@@ -69,6 +69,8 @@ typedef union VTD_IR_MSIAddress VTD_IR_MSIAddress;typedef struct VTDPASIDDirEntry VTDPASIDDirEntry ;typedef struct VTDPASIDEntry VTDPASIDEntry ;typedef struct VTDIOMMUContext VTDIOMMUContext ; +typedef struct VTDPASIDCacheEntry VTDPASIDCacheEntry ; +typedef struct VTDPASIDAddressSpace VTDPASIDAddressSpace ;/* Context-Entry */ struct VTDContextEntry {@@ -101,6 +103,31 @@ struct VTDPASIDEntry {uint64_t val[8 ]; }; +struct pasid_key { +uint32_t pasid; +uint16_t sid; +}; + +struct VTDPASIDCacheEntry { +/* +* The cache entry is obsolete if +* pasid_cache_gen!=IntelIOMMUState.pasid_cache_gen +*/ +uint32_t pasid_cache_gen; +struct VTDPASIDEntry pasid_entry; +};
include/hw/i386/intel_iommu.h struct VTDPASIDEntry { · line 128
Quoted 10 lines
include/hw/i386/intel_iommu.h +10
-0+ +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;
In vtd_pasid_cache_gsi() [1], you directly reset pasid_cache_gen for each pasid address space. You never increase pasid_cache_entry.pasid_cache_gen. Then IIUC the gen will always be either 0 or 1. And...
Quoted 14 lines
include/hw/i386/intel_iommu.h +3
-0+}; +struct VTDAddressSpace {PCIBus *bus;uint8_t devfn;@@ -122,6 +149,7 @@ struct VTDIOMMUContext {uint8_t devfn;IOMMUContext iommu_context;DualStageIOMMUObject *dsi_obj; +QLIST_ENTRY (VTDIOMMUContext) next;IntelIOMMUState *iommu_state; };
include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 304
Quoted 5 lines
include/hw/i386/intel_iommu.h +2
-0@@ -272,9 +300,14 @@ 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 */ +uint32_t pasid_cache_gen;/* Should be in [1,MAX] */
... This should always be 1.
IIUC you can drop both of the pasid_cache_gen, because in this whole patchset you'll remove the pasid hash entry when it is invalidated, right? Then if the hash entry is there, it must be valid. When it's out-dated, it'll be removed from the hash.
Quoted 22 lines
include/hw/i386/intel_iommu.h +5
-1/* list of registered notifiers */ QLIST_HEAD(, VTDAddressSpace) vtd_as_with_notifiers; +/* list of VTDIOMMUContexts with DualStageIOMMUObject registered */ + QLIST_HEAD(, VTDIOMMUContext) vtd_dev_icx_list; +/* interrupt remapping */ bool intr_enabled;/* Whether guest enabled IR */ dma_addr_t intr_root;/* Interrupt remapping table pointer */ @@ -291,7 +324,8 @@ struct IntelIOMMUState {/* * Protects IOMMU states in general. Currently it protects the -* per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace. +* per-IOMMU IOTLB cache, and context entry cache in VTDAddressSpace, +* and pasid cache in VTDPASIDAddressSpace. */ QemuMutex iommu_lock; };Signature · 1 line
2.7.4
Signature · 1 line
Peter Xu
RE: [RFC v3 16/25] intel_iommu: add PASID cache management infrastructure
Quoted 206 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 7:36 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 16/25]intel_iommu: add PASID cache management infrastructureQuoted 51 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>Quoted 2 lines
--- hw/i386/intel_iommu.c | 367+++++++++++++++++++++++++++++++++++++++++
Quoted 18 lines
hw/i386/intel_iommu_internal.h | 14 ++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 36 +++- 4 files changed, 417 insertions(+), 1 deletion(-)hw/i386/intel_iommu.c +1
-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) \hw/i386/intel_iommu.c line 67
Quoted 2 lines
hw/i386/intel_iommu.c +0
-0@@ -65,6 +66,8 @@static void vtd_address_space_refresh_all (IntelIOMMUState *s); static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier*n);
hw/i386/intel_iommu.c static inline bool vtd_pe_type_check(X86IOMMUState
Quoted 15 lines
hw/i386/intel_iommu.c +3
-0+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 +690,11 @@ static inline bool vtd_pe_type_check(X86IOMMUState
*x86_iommu,hw/i386/intel_iommu.c static bool
Quoted 12 lines
hw/i386/intel_iommu.c +5
-0return true; } +static inline uint16_t vtd_pe_get_domain_id (VTDPASIDEntry *pe) +{ +return VTD_SM_PASID_ENTRY_DID ((pe)->val[1 ]); +} +static inline bool vtd_pdire_present (VTDPASIDDirEntry *pdire) {return pdire->val &1 ;@@ -2393,19 +2402,370 @@ static bool
vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc)Quoted 53 lines
hw/i386/intel_iommu.c +50
-0return 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;hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc)
Quoted 10 lines
hw/i386/intel_iommu.c +10
-0+ } + + 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)
Quoted 31 lines
hw/i386/intel_iommu.c +31
-0+{ +return !memcmp (p1, p2,sizeof (*p1)); +} + +/** +* This function is used to clear pasid_cache_gen of 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; + +if (!(pc_entry->pasid_cache_gen == s->pasid_cache_gen)) { +return false; + } +
Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2501
Quoted 6 lines
hw/i386/intel_iommu.c +6
-0+switch (pc_info->flags &VTD_PASID_CACHE_INFO_MASK ) { +case VTD_PASID_CACHE_PASIDSI : +if (pc_info->pasid != pasid) { +return false; + } +/* Fall through */ Why fall through?
For VTD_PASID_CACHE_PASIDSI, it implies domain selective, so it requires to check did just as VTD_PASID_CACHE_DOMSI.
Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2506
Quoted 6 lines
hw/i386/intel_iommu.c +5
-0+case VTD_PASID_CACHE_DOMSI : +if (pc_info->domain_id != did) { +return false; + } +/* Fall through */ Same here.
If code comes to here, it means the necessary checks are passed. Should add a break here. However, as the below case does nothing and just calls break. So I let the code fall through.
Quoted 6 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2509
Quoted 4 lines
hw/i386/intel_iommu.c +3
-0+case VTD_PASID_CACHE_GLOBAL : +break ; +default :Nevee reach here right? If so we can abort.
yes, should never reach here.
Quoted 60 lines
Quoted 60 lines
hw/i386/intel_iommu.c +58
-0+return false; + } + +/* +* 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; + } +/* Compare cached pasid entry and latest pasid entry */ +if (!vtd_pasid_entry_compare (&pe, &pc_entry->pasid_entry)) { +/* pasid entry was updated, thus update the pasid cache */ + pc_entry->pasid_entry = pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +/* +* 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 false; +remove: +/* +* TODO: +* - send pasid unbind to host for passthru devices +* - when pasid-base-iotlb(piotlb) infrastructure is ready, +* should invalidate QEMU piotlb togehter with this change. +*/ +return true; +} +static int vtd_pasid_cache_dsi (IntelIOMMUState *s,uint16_t domain_id) { +VTDPASIDCacheInfo pc_info; + +trace_vtd_pasid_cache_dsi (domain_id); + + pc_info.flags =VTD_PASID_CACHE_DOMSI ; + pc_info.domain_id = domain_id; + +/* +* Loop all existing pasid caches and update them. +*/ +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + +/*
Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2574
Quoted 5 lines
hw/i386/intel_iommu.c +5
-0+* TODO: Domain selective PASID cache invalidation +* flushes all the pasid caches within a domain. To +* be safe, after invalidating the pasid caches, emulator +* needs to replay the pasid bindings by walking guest +* pasid dir and pasid table. Better spell out on what special case we're handling here: When the guest setup a new PASID entry then send a PASID DSI.
oh, yes. will add it in new version. :-)
Quoted 116 lines
Quoted 116 lines
hw/i386/intel_iommu.c +109
-0+*/ +vtd_iommu_unlock (s);return 0 ; } +/** +* 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, +bool allocate) +{ +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 && allocate) { + 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->context_cache_entry.context_cache_gen =0 ; + vtd_pasid_as->pasid = pasid; + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen =0 ; +g_hash_table_insert (s->vtd_pasid_as, new_key, vtd_pasid_as); + } +return vtd_pasid_as; +} + +/** +* This function updates the pasid entry cached in &vtd_pasid_as. +* Caller of this function should hold iommu_lock. +*/ +static inline void vtd_fill_in_pe_cache ( +VTDPASIDAddressSpace *vtd_pasid_as,VTDPASIDEntry *pe) +{ +IntelIOMMUState *s = vtd_pasid_as->iommu_state; +VTDPASIDCacheEntry *pc_entry = &vtd_pasid_as->pasid_cache_entry; + + pc_entry->pasid_entry = *pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +} +static int vtd_pasid_cache_psi (IntelIOMMUState *s,uint16_t domain_id,uint32_t pasid) { +VTDPASIDCacheInfo pc_info; +VTDPASIDEntry pe; +VTDBus *vtd_bus; +int bus_n, devfn; +VTDPASIDAddressSpace *vtd_pasid_as; +VTDIOMMUContext *vtd_icx; + +/* PASID selective implies a DID selective */ + pc_info.flags =VTD_PASID_CACHE_PASIDSI ; + pc_info.domain_id = domain_id; + pc_info.pasid = pasid; + +/* +* Regards to a pasid selective pasid cache invalidation (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 +* +* Here the handling of a PSI is: +* 1) loop all the exisitng 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) +* However, it is not good to always loop all devices. In this +* implementation. We do it in this ways: +* - For devices which have VTDIOMMUContext 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 VTDIOMMUContext +* instances, it is not necessary to create pasid cache at +* this phase since it could be created when vIOMMU do DMA +* address translation. This is not implemented yet since +* no PASID-capable emulated devices today. If we have it +* in future, the pasid cache shall be created there. +*/ + +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + + QLIST_FOREACH(vtd_icx, &s->vtd_dev_icx_list, next) { + vtd_bus = vtd_icx->vtd_bus;
Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2695
Quoted 6 lines
hw/i386/intel_iommu.c +6
-0+ devfn = vtd_icx->devfn; + bus_n =pci_bus_num (vtd_bus->bus); + +/* Step 1: fetch vtd_pasid_as and check if it is valid */ + vtd_pasid_as =vtd_add_find_pasid_as (s, vtd_bus, + devfn, pasid, true);I feel like you wanted to pass "false" here for "allocate".
emmm, yeah. It was "false" in draft code as step 1 is only to check if a valid vtd_pasid_as exists. And in step 3, it needs to call vtd_add_find_pasid_as() with "allocate" be "true". In vtd_add_find_pasid_as(), it will try search vtd_pasid_as first and then allocate a new one. In such logic, there will be two vtd_add_find_pasid_as() callig and means two hash table searching.
So I mofified it to be "true" to save a vtd_add_find_pasid_as() calling. If a vtd_pasid_as is valid, its pasid_cache_gen will be equal to s->pasid_cache_gen. If not, the vtd_pasid_as is a newly allocated and needs to go through step 2 and step 3 to fulfill it. Looks like I missed to free the vtd_pasid_as when step 2 failed. Will add it if you are fine with the current logic.
Quoted 232 lines
Quoted 54 lines
hw/i386/intel_iommu.c +51
-0+if (vtd_pasid_as && + (s->pasid_cache_gen == + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen)) { +/* +* pasid_cache_gen equals to s->pasid_cache_gen means +* vtd_pasid_as is valid after the above s->vtd_pasid_as +* updates. Thus no need for the below steps. +*/ +continue ; + } + +/* +* Step 2: vtd_pasid_as is not valid, it's potentailly a +* new pasid bind. Fetch guest pasid entry. +*/ +if (vtd_dev_get_pe_from_pasid (s, bus_n, devfn, pasid, &pe)) { +continue ; + } + +/* +* Step 3: pasid entry exists, update pasid cache +* +* Here need to check domain ID since guest pasid entry +* exists. What needs to do are: +* - update the pc_entry in the vtd_pasid_as +* - set proper pc_entry.pasid_cache_gen +* - pass down the latest guest pasid entry config to host +* (will be added in later patch) +*/ +if (domain_id ==vtd_pe_get_domain_id (&pe)) { +vtd_fill_in_pe_cache (vtd_pasid_as, &pe); + } + } +vtd_iommu_unlock (s);return 0 ; } +/** +* 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.flags =VTD_PASID_CACHE_GLOBAL ; + +/* +* Reset pasid cache is a big hammer, so use +* g_hash_table_foreach_remove which will free +* the vtd_pasid_as instances. +*/ + g_hash_table_foreach_remove(s->vtd_pasid_as,hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2759
Quoted 10 lines
hw/i386/intel_iommu.c +8
-0+ vtd_flush_pasid, &pc_info); + s->pasid_cache_gen =1 ; +} +static int vtd_pasid_cache_gsi (IntelIOMMUState *s) { +trace_vtd_pasid_cache_gsi (); + +vtd_iommu_lock (s); +vtd_pasid_cache_reset (s);[1]
hw/i386/intel_iommu.c static int
Quoted 13 lines
hw/i386/intel_iommu.c +9
-0+ +/* +* TODO: Global PASID cache invalidation may be +* flushes all the pasid caches. To be safe, after +* invalidating the pasid caches, emulator needs +* to replay the pasid bindings by walking guest +* pasid dir and pasid table. +*/ +vtd_iommu_unlock (s);return 0 ; }@@ -3659,8 +4019,11 @@ static intvtd_icx_register_ds_iommu(IOMMUContext
*iommu_ctx,hw/i386/intel_iommu.c static void
Quoted 12 lines
hw/i386/intel_iommu.c +3
-0VTDIOMMUContext *vtd_dev_icx =container_of (iommu_ctx, VTDIOMMUContext, iommu_context); +IntelIOMMUState *s = vtd_dev_icx->iommu_state; vtd_dev_icx->dsi_obj = dsi_obj; +QLIST_INSERT_HEAD (&s->vtd_dev_icx_list, vtd_dev_icx, next); +return 0 ; }@@ -3672,6 +4035,7 @@ static voidvtd_icx_unregister_ds_iommu(IOMMUContext
*iommu_ctx,Quoted 16 lines
hw/i386/intel_iommu.c +2
-0iommu_context); vtd_dev_icx->dsi_obj =NULL ; +QLIST_REMOVE (vtd_dev_icx, next); }IOMMUContextOps vtd_iommu_context_ops = {@@ -4130,6 +4494,7 @@ 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.c static void vtd_realize(DeviceState *dev, Error **errp) · line 4521
Quoted 2 lines
hw/i386/intel_iommu.c +0
-0@@ -4155,6 +4520,8 @@ 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,hw/i386/intel_iommu.c static void vtd_realize(DeviceState *dev, Error **errp)
Quoted 5 lines
hw/i386/intel_iommu.c +2
-0g_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);hw/i386/intel_iommu_internal.h typedef enum VTDFaultReason { · line 314
Quoted 10 lines
hw/i386/intel_iommu.c +0
-0pci_setup_iommu (bus, &vtd_iommu_ops, dev);hw/i386/intel_iommu_internal.h +1
-0@@ -311,6 +311,7 @@ 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*/
Quoted 28 lines
hw/i386/intel_iommu_internal.h +13
-0/* This is not a normal fault reason. We use this to indicate some faults * that are not referenced by the VT-d specification. @@ -485,6 +486,19 @@ struct VTDRootEntry { };typedef struct VTDRootEntry VTDRootEntry ; +struct VTDPASIDCacheInfo { +#define VTD_PASID_CACHE_GLOBAL (1ULL << 0) +#define VTD_PASID_CACHE_DOMSI (1ULL << 1) +#define VTD_PASID_CACHE_PASIDSI (1ULL << 2) +uint32_t flags; +uint16_t domain_id; +uint32_t pasid; +}; +#define VTD_PASID_CACHE_INFO_MASK (VTD_PASID_CACHE_GLOBAL | \ + VTD_PASID_CACHE_DOMSI | \ + VTD_PASID_CACHE_PASIDSI) +typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ; +/* Masks for struct VTDRootEntry */ #define VTD_ROOT_ENTRY_P 1ULL#define VTD_ROOT_ENTRY_CTP (~0xfffULL)hw/i386/trace-events +0
-0hw/i386/trace-events vtd_inv_qi_head(uint16_t head) "read head %d" · line 27
Quoted 6 lines
hw/i386/trace-events +1
-0@@ -22,6 +22,7 @@ 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_reset(void) "" vtd_pasid_cache_gsi(void) "" vtd_pasid_cache_dsi(uint16_t domain) "Domian slective PC invalidationdomain 0x%"PRIx16
hw/i386/trace-events vtd_inv_qi_head(uint16_t head) "read head %d" · line 28
Quoted 1 line
hw/i386/trace-events +0
-0vtd_pasid_cache_psi(uint16_t domain, uint32_t pasid) "PASID slective PCinvalidation domain 0x%"PRIx16" pasid 0x%"PRIx32
Quoted 30 lines
include/hw/i386/intel_iommu.h +15
-0@@ -69,6 +69,8 @@ typedef union VTD_IR_MSIAddress VTD_IR_MSIAddress;typedef struct VTDPASIDDirEntry VTDPASIDDirEntry ;typedef struct VTDPASIDEntry VTDPASIDEntry ;typedef struct VTDIOMMUContext VTDIOMMUContext ; +typedef struct VTDPASIDCacheEntry VTDPASIDCacheEntry ; +typedef struct VTDPASIDAddressSpace VTDPASIDAddressSpace ;/* Context-Entry */ struct VTDContextEntry {@@ -101,6 +103,31 @@ struct VTDPASIDEntry {uint64_t val[8 ]; }; +struct pasid_key { +uint32_t pasid; +uint16_t sid; +}; + +struct VTDPASIDCacheEntry { +/* +* The cache entry is obsolete if +* pasid_cache_gen!=IntelIOMMUState.pasid_cache_gen +*/ +uint32_t pasid_cache_gen; +struct VTDPASIDEntry pasid_entry; +};include/hw/i386/intel_iommu.h struct VTDPASIDEntry { · line 128
Quoted 10 lines
include/hw/i386/intel_iommu.h +10
-0+ +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;In
vtd_pasid_cache_gsi()[1], you directly resetpasid_cache_genfor each pasid address space. You never increase pasid_cache_entry.pasid_cache_gen. Then IIUC the gen will always be either 0 or 1. And...Quoted 14 lines
include/hw/i386/intel_iommu.h +3
-0+}; +struct VTDAddressSpace {PCIBus *bus;uint8_t devfn;@@ -122,6 +149,7 @@ struct VTDIOMMUContext {uint8_t devfn;IOMMUContext iommu_context;DualStageIOMMUObject *dsi_obj; +QLIST_ENTRY (VTDIOMMUContext) next;IntelIOMMUState *iommu_state; };include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 301
Quoted 2 lines
include/hw/i386/intel_iommu.h +0
-0@@ -272,9 +300,14 @@ struct IntelIOMMUState {GHashTable *vtd_as_by_busptr;/* VTDBus objects indexed by PCIBus* reference */
include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 302
Quoted 1 line
include/hw/i386/intel_iommu.h +0
-0VTDBus *vtd_as_by_bus_num[VTD_PCI_BUS_MAX ];/* VTDBus objects indexed by bus number */
Quoted 8 lines
include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 304
Quoted 2 lines
include/hw/i386/intel_iommu.h +2
-0+GHashTable *vtd_pasid_as;/* VTDPASIDAddressSpace instances */ +uint32_t pasid_cache_gen;/* Should be in [1,MAX] */ ... This should always be 1.
IIUC you can drop both of thepasid_cache_gen, because in this whole patchset you'll remove the pasid hash entry when it is invalidated, right? Then if the hash entry is there, it must be valid. When it's out-dated, it'll be removed from the hash.
Oh, yes it is. However, it's not my intetion. I'd like to let [1] to increase the s->pasid_cache_gen instead of justing zero it. I think it will save some time as loop hash table takes time. Thanks for catching it. :-)
Regards,
Yi Liu
Re: [RFC v3 16/25] intel_iommu: add PASID cache management infrastructure
Quoted 215 lines · Yi L
Quoted 206 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 7:36 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 16/25]intel_iommu: add PASID cache management infrastructureQuoted 51 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>Quoted 2 lines
--- hw/i386/intel_iommu.c | 367+++++++++++++++++++++++++++++++++++++++++
Quoted 18 lines
hw/i386/intel_iommu_internal.h | 14 ++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 36 +++- 4 files changed, 417 insertions(+), 1 deletion(-)hw/i386/intel_iommu.c +1
-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) \hw/i386/intel_iommu.c line 67
Quoted 2 lines
hw/i386/intel_iommu.c +0
-0@@ -65,6 +66,8 @@static void vtd_address_space_refresh_all (IntelIOMMUState *s); static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier*n);
hw/i386/intel_iommu.c static inline bool vtd_pe_type_check(X86IOMMUState
Quoted 15 lines
hw/i386/intel_iommu.c +3
-0+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 +690,11 @@ static inline bool vtd_pe_type_check(X86IOMMUState
*x86_iommu,hw/i386/intel_iommu.c static bool
Quoted 12 lines
hw/i386/intel_iommu.c +5
-0return true; } +static inline uint16_t vtd_pe_get_domain_id (VTDPASIDEntry *pe) +{ +return VTD_SM_PASID_ENTRY_DID ((pe)->val[1 ]); +} +static inline bool vtd_pdire_present (VTDPASIDDirEntry *pdire) {return pdire->val &1 ;@@ -2393,19 +2402,370 @@ static bool
vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc)Quoted 53 lines
hw/i386/intel_iommu.c +50
-0return 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;hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc)
Quoted 10 lines
hw/i386/intel_iommu.c +10
-0+ } + + 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)
Quoted 31 lines
hw/i386/intel_iommu.c +31
-0+{ +return !memcmp (p1, p2,sizeof (*p1)); +} + +/** +* This function is used to clear pasid_cache_gen of 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; + +if (!(pc_entry->pasid_cache_gen == s->pasid_cache_gen)) { +return false; + } +Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2501
Quoted 6 lines
hw/i386/intel_iommu.c +6
-0+switch (pc_info->flags &VTD_PASID_CACHE_INFO_MASK ) { +case VTD_PASID_CACHE_PASIDSI : +if (pc_info->pasid != pasid) { +return false; + } +/* Fall through */ Why fall through?
Quoted 2 lines
For
VTD_PASID_CACHE_PASIDSI, it implies domain selective, so it requires to check did just asVTD_PASID_CACHE_DOMSI.
Ah right. :)
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
hw/i386/intel_iommu.c +5
-0+case VTD_PASID_CACHE_DOMSI : +if (pc_info->domain_id != did) { +return false; + } +/* Fall through */
Quoted 5 lines
Quoted 1 line
Same here.
If code comes to here, it means the necessary checks are passed. Should add a break here. However, as the below case does nothing and just calls break. So I let the code fall through.
Yes this is fine too.
Quoted 213 lines
Quoted 7 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2509
Quoted 5 lines
hw/i386/intel_iommu.c +3
-0+case VTD_PASID_CACHE_GLOBAL : +break ; +default :Nevee reach here right? If so we can abort.
yes, should never reach here.
Quoted 60 lines
Quoted 60 lines
hw/i386/intel_iommu.c +58
-0+return false; + } + +/* +* 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; + } +/* Compare cached pasid entry and latest pasid entry */ +if (!vtd_pasid_entry_compare (&pe, &pc_entry->pasid_entry)) { +/* pasid entry was updated, thus update the pasid cache */ + pc_entry->pasid_entry = pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +/* +* 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 false; +remove: +/* +* TODO: +* - send pasid unbind to host for passthru devices +* - when pasid-base-iotlb(piotlb) infrastructure is ready, +* should invalidate QEMU piotlb togehter with this change. +*/ +return true; +} +static int vtd_pasid_cache_dsi (IntelIOMMUState *s,uint16_t domain_id) { +VTDPASIDCacheInfo pc_info; + +trace_vtd_pasid_cache_dsi (domain_id); + + pc_info.flags =VTD_PASID_CACHE_DOMSI ; + pc_info.domain_id = domain_id; + +/* +* Loop all existing pasid caches and update them. +*/ +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + +/* Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2574
Quoted 5 lines
hw/i386/intel_iommu.c +5
-0+* TODO: Domain selective PASID cache invalidation +* flushes all the pasid caches within a domain. To +* be safe, after invalidating the pasid caches, emulator +* needs to replay the pasid bindings by walking guest +* pasid dir and pasid table. Better spell out on what special case we're handling here: When the guest setup a new PASID entry then send a PASID DSI.
oh, yes. will add it in new version. :-)
Quoted 116 lines
Quoted 116 lines
hw/i386/intel_iommu.c +109
-0+*/ +vtd_iommu_unlock (s);return 0 ; } +/** +* 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, +bool allocate) +{ +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 && allocate) { + 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->context_cache_entry.context_cache_gen =0 ; + vtd_pasid_as->pasid = pasid; + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen =0 ; +g_hash_table_insert (s->vtd_pasid_as, new_key, vtd_pasid_as); + } +return vtd_pasid_as; +} + +/** +* This function updates the pasid entry cached in &vtd_pasid_as. +* Caller of this function should hold iommu_lock. +*/ +static inline void vtd_fill_in_pe_cache ( +VTDPASIDAddressSpace *vtd_pasid_as,VTDPASIDEntry *pe) +{ +IntelIOMMUState *s = vtd_pasid_as->iommu_state; +VTDPASIDCacheEntry *pc_entry = &vtd_pasid_as->pasid_cache_entry; + + pc_entry->pasid_entry = *pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +} +static int vtd_pasid_cache_psi (IntelIOMMUState *s,uint16_t domain_id,uint32_t pasid) { +VTDPASIDCacheInfo pc_info; +VTDPASIDEntry pe; +VTDBus *vtd_bus; +int bus_n, devfn; +VTDPASIDAddressSpace *vtd_pasid_as; +VTDIOMMUContext *vtd_icx; + +/* PASID selective implies a DID selective */ + pc_info.flags =VTD_PASID_CACHE_PASIDSI ; + pc_info.domain_id = domain_id; + pc_info.pasid = pasid; + +/* +* Regards to a pasid selective pasid cache invalidation (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 +* +* Here the handling of a PSI is: +* 1) loop all the exisitng 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) +* However, it is not good to always loop all devices. In this +* implementation. We do it in this ways: +* - For devices which have VTDIOMMUContext 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 VTDIOMMUContext +* instances, it is not necessary to create pasid cache at +* this phase since it could be created when vIOMMU do DMA +* address translation. This is not implemented yet since +* no PASID-capable emulated devices today. If we have it +* in future, the pasid cache shall be created there. +*/ + +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + + QLIST_FOREACH(vtd_icx, &s->vtd_dev_icx_list, next) { + vtd_bus = vtd_icx->vtd_bus;Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2695
Quoted 6 lines
hw/i386/intel_iommu.c +6
-0+ devfn = vtd_icx->devfn; + bus_n =pci_bus_num (vtd_bus->bus); + +/* Step 1: fetch vtd_pasid_as and check if it is valid */ + vtd_pasid_as =vtd_add_find_pasid_as (s, vtd_bus, + devfn, pasid, true);I feel like you wanted to pass "false" here for "allocate".
emmm, yeah. It was "false" in draft code as step 1 is only to check if a valid
vtd_pasid_asexists. And in step 3, it needs to callvtd_add_find_pasid_as()with "allocate" be "true". Invtd_add_find_pasid_as(), it will try searchvtd_pasid_asfirst and then allocate a new one. In such logic, there will be twovtd_add_find_pasid_as()callig and means two hash table searching.
Quoted 5 lines
So I mofified it to be "true" to save a
vtd_add_find_pasid_as()calling. If avtd_pasid_asis valid, itspasid_cache_genwill be equal tos->pasid_cache_gen. If not, thevtd_pasid_asis a newly allocated and needs to go through step 2 and step 3 to fulfill it. Looks like I missed to free thevtd_pasid_aswhen step 2 failed. Will add it if you are fine with the current logic.
I see. Note that vtd_add_find_pasid_as() is fast for no allocation, because hash lookup is O(1). However I think current approach is ok, but if with that, we can also:
- Remove the allocate parameter for vtd_add_find_pasid_as(), since it's always true even in future patches so useless,
- Remove the vtd_pasid_as check right below because it's not needed.
Quoted 3 lines
Quoted 3 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2696
Quoted 3 lines
hw/i386/intel_iommu.c +1
-0+ if (vtd_pasid_as &&
^^^^^^^^^^^^
Quoted 240 lines
Quoted 231 lines
Quoted 53 lines
hw/i386/intel_iommu.c +50
-0+ (s->pasid_cache_gen == + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen)) { +/* +* pasid_cache_gen equals to s->pasid_cache_gen means +* vtd_pasid_as is valid after the above s->vtd_pasid_as +* updates. Thus no need for the below steps. +*/ +continue ; + } + +/* +* Step 2: vtd_pasid_as is not valid, it's potentailly a +* new pasid bind. Fetch guest pasid entry. +*/ +if (vtd_dev_get_pe_from_pasid (s, bus_n, devfn, pasid, &pe)) { +continue ; + } + +/* +* Step 3: pasid entry exists, update pasid cache +* +* Here need to check domain ID since guest pasid entry +* exists. What needs to do are: +* - update the pc_entry in the vtd_pasid_as +* - set proper pc_entry.pasid_cache_gen +* - pass down the latest guest pasid entry config to host +* (will be added in later patch) +*/ +if (domain_id ==vtd_pe_get_domain_id (&pe)) { +vtd_fill_in_pe_cache (vtd_pasid_as, &pe); + } + } +vtd_iommu_unlock (s);return 0 ; } +/** +* 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.flags =VTD_PASID_CACHE_GLOBAL ; + +/* +* Reset pasid cache is a big hammer, so use +* g_hash_table_foreach_remove which will free +* the vtd_pasid_as instances. +*/ + g_hash_table_foreach_remove(s->vtd_pasid_as,hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2759
Quoted 10 lines
hw/i386/intel_iommu.c +8
-0+ vtd_flush_pasid, &pc_info); + s->pasid_cache_gen =1 ; +} +static int vtd_pasid_cache_gsi (IntelIOMMUState *s) { +trace_vtd_pasid_cache_gsi (); + +vtd_iommu_lock (s); +vtd_pasid_cache_reset (s);[1]
hw/i386/intel_iommu.c static int
Quoted 13 lines
hw/i386/intel_iommu.c +9
-0+ +/* +* TODO: Global PASID cache invalidation may be +* flushes all the pasid caches. To be safe, after +* invalidating the pasid caches, emulator needs +* to replay the pasid bindings by walking guest +* pasid dir and pasid table. +*/ +vtd_iommu_unlock (s);return 0 ; }@@ -3659,8 +4019,11 @@ static intvtd_icx_register_ds_iommu(IOMMUContext
*iommu_ctx,hw/i386/intel_iommu.c static void
Quoted 12 lines
hw/i386/intel_iommu.c +3
-0VTDIOMMUContext *vtd_dev_icx =container_of (iommu_ctx, VTDIOMMUContext, iommu_context); +IntelIOMMUState *s = vtd_dev_icx->iommu_state; vtd_dev_icx->dsi_obj = dsi_obj; +QLIST_INSERT_HEAD (&s->vtd_dev_icx_list, vtd_dev_icx, next); +return 0 ; }@@ -3672,6 +4035,7 @@ static voidvtd_icx_unregister_ds_iommu(IOMMUContext
*iommu_ctx,Quoted 16 lines
hw/i386/intel_iommu.c +2
-0iommu_context); vtd_dev_icx->dsi_obj =NULL ; +QLIST_REMOVE (vtd_dev_icx, next); }IOMMUContextOps vtd_iommu_context_ops = {@@ -4130,6 +4494,7 @@ 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.c static void vtd_realize(DeviceState *dev, Error **errp) · line 4521
Quoted 2 lines
hw/i386/intel_iommu.c +0
-0@@ -4155,6 +4520,8 @@ 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,hw/i386/intel_iommu.c static void vtd_realize(DeviceState *dev, Error **errp)
Quoted 5 lines
hw/i386/intel_iommu.c +2
-0g_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);hw/i386/intel_iommu_internal.h typedef enum VTDFaultReason { · line 314
Quoted 10 lines
hw/i386/intel_iommu.c +0
-0pci_setup_iommu (bus, &vtd_iommu_ops, dev);hw/i386/intel_iommu_internal.h +1
-0@@ -311,6 +311,7 @@ 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*/
Quoted 28 lines
hw/i386/intel_iommu_internal.h +13
-0/* This is not a normal fault reason. We use this to indicate some faults * that are not referenced by the VT-d specification. @@ -485,6 +486,19 @@ struct VTDRootEntry { };typedef struct VTDRootEntry VTDRootEntry ; +struct VTDPASIDCacheInfo { +#define VTD_PASID_CACHE_GLOBAL (1ULL << 0) +#define VTD_PASID_CACHE_DOMSI (1ULL << 1) +#define VTD_PASID_CACHE_PASIDSI (1ULL << 2) +uint32_t flags; +uint16_t domain_id; +uint32_t pasid; +}; +#define VTD_PASID_CACHE_INFO_MASK (VTD_PASID_CACHE_GLOBAL | \ + VTD_PASID_CACHE_DOMSI | \ + VTD_PASID_CACHE_PASIDSI) +typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ; +/* Masks for struct VTDRootEntry */ #define VTD_ROOT_ENTRY_P 1ULL#define VTD_ROOT_ENTRY_CTP (~0xfffULL)hw/i386/trace-events +0
-0hw/i386/trace-events vtd_inv_qi_head(uint16_t head) "read head %d" · line 27
Quoted 6 lines
hw/i386/trace-events +1
-0@@ -22,6 +22,7 @@ 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_reset(void) "" vtd_pasid_cache_gsi(void) "" vtd_pasid_cache_dsi(uint16_t domain) "Domian slective PC invalidationdomain 0x%"PRIx16
hw/i386/trace-events vtd_inv_qi_head(uint16_t head) "read head %d" · line 28
Quoted 1 line
hw/i386/trace-events +0
-0vtd_pasid_cache_psi(uint16_t domain, uint32_t pasid) "PASID slective PCinvalidation domain 0x%"PRIx16" pasid 0x%"PRIx32
Quoted 30 lines
include/hw/i386/intel_iommu.h +15
-0@@ -69,6 +69,8 @@ typedef union VTD_IR_MSIAddress VTD_IR_MSIAddress;typedef struct VTDPASIDDirEntry VTDPASIDDirEntry ;typedef struct VTDPASIDEntry VTDPASIDEntry ;typedef struct VTDIOMMUContext VTDIOMMUContext ; +typedef struct VTDPASIDCacheEntry VTDPASIDCacheEntry ; +typedef struct VTDPASIDAddressSpace VTDPASIDAddressSpace ;/* Context-Entry */ struct VTDContextEntry {@@ -101,6 +103,31 @@ struct VTDPASIDEntry {uint64_t val[8 ]; }; +struct pasid_key { +uint32_t pasid; +uint16_t sid; +}; + +struct VTDPASIDCacheEntry { +/* +* The cache entry is obsolete if +* pasid_cache_gen!=IntelIOMMUState.pasid_cache_gen +*/ +uint32_t pasid_cache_gen; +struct VTDPASIDEntry pasid_entry; +};include/hw/i386/intel_iommu.h struct VTDPASIDEntry { · line 128
Quoted 10 lines
include/hw/i386/intel_iommu.h +10
-0+ +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;In
vtd_pasid_cache_gsi()[1], you directly resetpasid_cache_genfor each pasid address space. You never increase pasid_cache_entry.pasid_cache_gen. Then IIUC the gen will always be either 0 or 1. And...Quoted 14 lines
include/hw/i386/intel_iommu.h +3
-0+}; +struct VTDAddressSpace {PCIBus *bus;uint8_t devfn;@@ -122,6 +149,7 @@ struct VTDIOMMUContext {uint8_t devfn;IOMMUContext iommu_context;DualStageIOMMUObject *dsi_obj; +QLIST_ENTRY (VTDIOMMUContext) next;IntelIOMMUState *iommu_state; };include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 301
Quoted 2 lines
include/hw/i386/intel_iommu.h +0
-0@@ -272,9 +300,14 @@ struct IntelIOMMUState {GHashTable *vtd_as_by_busptr;/* VTDBus objects indexed by PCIBus* reference */
include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 302
Quoted 1 line
include/hw/i386/intel_iommu.h +0
-0VTDBus *vtd_as_by_bus_num[VTD_PCI_BUS_MAX ];/* VTDBus objects indexed by bus number */
Quoted 8 lines
include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 304
Quoted 2 lines
include/hw/i386/intel_iommu.h +2
-0+GHashTable *vtd_pasid_as;/* VTDPASIDAddressSpace instances */ +uint32_t pasid_cache_gen;/* Should be in [1,MAX] */ ... This should always be 1.
IIUC you can drop both of thepasid_cache_gen, because in this whole patchset you'll remove the pasid hash entry when it is invalidated, right? Then if the hash entry is there, it must be valid. When it's out-dated, it'll be removed from the hash.
Quoted 4 lines
Oh, yes it is. However, it's not my intetion. I'd like to let [1] to increase the
s->pasid_cache_geninstead of justing zero it. I think it will save some time as loop hash table takes time. Thanks for catching it. :-)
OK that's fine too. Then remember to conditionally reset it:
static int vtd_pasid_cache_gsi(IntelIOMMUState *s)
{
trace_vtd_pasid_cache_gsi();
vtd_iommu_lock(s);
s->pasid_cache_gen++;
if (s->pasid_cache_gen >= THRESHOLD) {
vtd_pasid_cache_reset(s);
}
vtd_iommu_unlock(s);
return 0;
}Thanks,
Signature · 1 line
Peter Xu
RE: [RFC v3 16/25] intel_iommu: add PASID cache management infrastructure
Quoted 90 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 11:26 PM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 16/25]intel_iommu: add PASID cache management infrastructureQuoted 83 lines · Yi L
Quoted 83 lines
From: Peter Xu <[email protected]>
Sent: Wednesday, February 12, 2020 7:36 AM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 16/25]intel_iommu: add PASID cache management infrastructureQuoted 51 lines · Yi L
From: Liu Yi L <[email protected]>
This 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.
Cc Kevin Tian <[email protected]>Cc Jacob Pan <[email protected]>Cc Peter Xu <[email protected]>Cc Yi Sun <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Richard Henderson <[email protected]>Cc Eduardo Habkost <[email protected]>Signed-off-by Liu Yi L <[email protected]>Quoted 2 lines
--- hw/i386/intel_iommu.c | 367+++++++++++++++++++++++++++++++++++++++++
Quoted 17 lines
hw/i386/intel_iommu_internal.h | 14 ++ hw/i386/trace-events | 1 + include/hw/i386/intel_iommu.h | 36 +++- 4 files changed, 417 insertions(+), 1 deletion(-)hw/i386/intel_iommu.c index renamed from hw/i386/intel_iommu.c +0
-058e7213..c75cb7b 100644
hw/i386/intel_iommu.c +1
-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 voidQuoted 2 lines
vtd_address_space_refresh_all(IntelIOMMUState *s); static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier*n);
Quoted 2 lines
+static void
vtd_pasid_cache_reset(IntelIOMMUState *s);
[...]
Quoted 282 lines
Quoted 37 lines
Quoted 28 lines
Quoted 28 lines
hw/i386/intel_iommu.c +28
-0+ +/** +* This function is used to clear pasid_cache_gen of 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; + +if (!(pc_entry->pasid_cache_gen == s->pasid_cache_gen)) { +return false; + } +Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2501
Quoted 6 lines
hw/i386/intel_iommu.c +6
-0+switch (pc_info->flags &VTD_PASID_CACHE_INFO_MASK ) { +case VTD_PASID_CACHE_PASIDSI : +if (pc_info->pasid != pasid) { +return false; + } +/* Fall through */ Why fall through?
Quoted 2 lines
For
VTD_PASID_CACHE_PASIDSI, it implies domain selective, so it requires to check did just asVTD_PASID_CACHE_DOMSI.Ah right. :)
Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
hw/i386/intel_iommu.c +5
-0+case VTD_PASID_CACHE_DOMSI : +if (pc_info->domain_id != did) { +return false; + } +/* Fall through */ Quoted 5 lines
Quoted 1 line
Same here.
If code comes to here, it means the necessary checks are passed. Should add a break here. However, as the below case does nothing and just calls break. So I let the code fall through.
Yes this is fine too.
Quoted 87 lines
Quoted 7 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2509
Quoted 5 lines
hw/i386/intel_iommu.c +3
-0+case VTD_PASID_CACHE_GLOBAL : +break ; +default :Nevee reach here right? If so we can abort.
yes, should never reach here.
Quoted 60 lines
Quoted 60 lines
hw/i386/intel_iommu.c +43
-0+return false; + } + +/* +* 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; + } +/* Compare cached pasid entry and latest pasid entry */ +if (!vtd_pasid_entry_compare (&pe, &pc_entry->pasid_entry)) { +/* pasid entry was updated, thus update the pasid cache */ + pc_entry->pasid_entry = pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; +/* +* 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 false; +remove: +/* +* TODO: +* - send pasid unbind to host for passthru devices +* - when pasid-base-iotlb(piotlb) infrastructure is ready, +* should invalidate QEMU piotlb togehter with this change. +*/ +return true; +} + static int vtd_pasid_cache_dsi(IntelIOMMUState *s, uint16_t
domain_id) {hw/i386/intel_iommu.c +15
-0+VTDPASIDCacheInfo pc_info; + +trace_vtd_pasid_cache_dsi (domain_id); + + pc_info.flags =VTD_PASID_CACHE_DOMSI ; + pc_info.domain_id = domain_id; + +/* +* Loop all existing pasid caches and update them. +*/ +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + +/* Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2574
Quoted 5 lines
hw/i386/intel_iommu.c +5
-0+* TODO: Domain selective PASID cache invalidation +* flushes all the pasid caches within a domain. To +* be safe, after invalidating the pasid caches, emulator +* needs to replay the pasid bindings by walking guest +* pasid dir and pasid table. Better spell out on what special case we're handling here: When the guest setup a new PASID entry then send a PASID DSI.
oh, yes. will add it in new version. :-)
Quoted 6 lines
Quoted 6 lines
hw/i386/intel_iommu.c +2
-0+*/ +vtd_iommu_unlock (s);return 0 ; }Quoted 8 lines
Quoted 8 lines
hw/i386/intel_iommu.c
Quoted 8 lines
hw/i386/intel_iommu.c +8
-0+/** +* 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,
Quoted 120 lines
Quoted 104 lines
Quoted 104 lines
hw/i386/intel_iommu.c +53
-0+VTDBus *vtd_bus, + int devfn, + uint32_t pasid, + bool allocate) +{ + 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 && allocate) { + 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->context_cache_entry.context_cache_gen =0 ; + vtd_pasid_as->pasid = pasid; + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen =0 ; +g_hash_table_insert (s->vtd_pasid_as, new_key, vtd_pasid_as); + } +return vtd_pasid_as; +} + +/** +* This function updates the pasid entry cached in &vtd_pasid_as. +* Caller of this function should hold iommu_lock. +*/ +static inline void vtd_fill_in_pe_cache ( +VTDPASIDAddressSpace *vtd_pasid_as,VTDPASIDEntry +*pe) { +IntelIOMMUState *s = vtd_pasid_as->iommu_state; +VTDPASIDCacheEntry *pc_entry = +&vtd_pasid_as->pasid_cache_entry; + + pc_entry->pasid_entry = *pe; + pc_entry->pasid_cache_gen = s->pasid_cache_gen; } + static int vtd_pasid_cache_psi(IntelIOMMUState *s, uint16_t domain_id, uint32_tpasid) {
hw/i386/intel_iommu.c +48
-0+VTDPASIDCacheInfo pc_info; +VTDPASIDEntry pe; +VTDBus *vtd_bus; +int bus_n, devfn; +VTDPASIDAddressSpace *vtd_pasid_as; +VTDIOMMUContext *vtd_icx; + +/* PASID selective implies a DID selective */ + pc_info.flags =VTD_PASID_CACHE_PASIDSI ; + pc_info.domain_id = domain_id; + pc_info.pasid = pasid; + +/* +* Regards to a pasid selective pasid cache invalidation (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 +* +* Here the handling of a PSI is: +* 1) loop all the exisitng 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) +* However, it is not good to always loop all devices. In this +* implementation. We do it in this ways: +* - For devices which have VTDIOMMUContext 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 VTDIOMMUContext +* instances, it is not necessary to create pasid cache at +* this phase since it could be created when vIOMMU do DMA +* address translation. This is not implemented yet since +* no PASID-capable emulated devices today. If we have it +* in future, the pasid cache shall be created there. +*/ + +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + + QLIST_FOREACH(vtd_icx, &s->vtd_dev_icx_list, next) { + vtd_bus = vtd_icx->vtd_bus;Quoted 8 lines
hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2695
Quoted 6 lines
hw/i386/intel_iommu.c +6
-0+ devfn = vtd_icx->devfn; + bus_n =pci_bus_num (vtd_bus->bus); + +/* Step 1: fetch vtd_pasid_as and check if it is valid */ + vtd_pasid_as =vtd_add_find_pasid_as (s, vtd_bus, + devfn, pasid, true);I feel like you wanted to pass "false" here for "allocate".
emmm, yeah. It was "false" in draft code as step 1 is only to check if a valid
vtd_pasid_asexists. And in step 3, it needs to callvtd_add_find_pasid_as()with "allocate" be "true". Invtd_add_find_pasid_as(), it will try searchvtd_pasid_asfirst and then allocate a new one. In such logic, there will be twovtd_add_find_pasid_as()callig and means two hash table searching.Quoted 7 lines
So I mofified it to be "true" to save a
vtd_add_find_pasid_as()calling. If avtd_pasid_asis valid, itspasid_cache_genwill be equal to s-pasid_cache_gen.
If not, thevtd_pasid_asis a newly allocated and needs to go through step 2 and step 3 to fulfill it. Looks like I missed to free thevtd_pasid_aswhen step
2 failed. Will add it if you are fine with the current logic.
Quoted 6 lines
I see. Note that
vtd_add_find_pasid_as()is fast for no allocation, because hash lookup is O(1). However I think current approach is ok, but if with that, we can also:- Remove the allocate parameter for
vtd_add_find_pasid_as(), since it's always true even in future patches so useless,
right. :-)
Quoted 6 lines
- Remove the
vtd_pasid_ascheck right below because it's not needed.Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
+ if (
vtd_pasid_as&&^^^^^^^^^^^^
yes, it is. In current series vtd_add_find_pasid_as() doesn’t check the result of vtd_pasid_as mem allocation, so no need to check vtd_pasid_as here either. However, it might be better to check the allocation result or it will result in issue if allocation failed. What's your preference here?
Quoted 75 lines
Quoted 75 lines
Quoted 75 lines
Quoted 50 lines
hw/i386/intel_iommu.c +47
-0+ (s->pasid_cache_gen == + vtd_pasid_as->pasid_cache_entry.pasid_cache_gen)) { +/* +* pasid_cache_gen equals to s->pasid_cache_gen means +* vtd_pasid_as is valid after the above s->vtd_pasid_as +* updates. Thus no need for the below steps. +*/ +continue ; + } + +/* +* Step 2: vtd_pasid_as is not valid, it's potentailly a +* new pasid bind. Fetch guest pasid entry. +*/ +if (vtd_dev_get_pe_from_pasid (s, bus_n, devfn, pasid, &pe)) { +continue ; + } + +/* +* Step 3: pasid entry exists, update pasid cache +* +* Here need to check domain ID since guest pasid entry +* exists. What needs to do are: +* - update the pc_entry in the vtd_pasid_as +* - set proper pc_entry.pasid_cache_gen +* - pass down the latest guest pasid entry config to host +* (will be added in later patch) +*/ +if (domain_id ==vtd_pe_get_domain_id (&pe)) { +vtd_fill_in_pe_cache (vtd_pasid_as, &pe); + } + } +vtd_iommu_unlock (s);return 0 ; } +/** +* 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.flags =VTD_PASID_CACHE_GLOBAL ; + +/* +* Reset pasid cache is a big hammer, so use +* g_hash_table_foreach_remove which will free +* the vtd_pasid_as instances. +*/ hw/i386/intel_iommu.c static bool vtd_process_iotlb_desc(IntelIOMMUState *s, VTDInvDesc *inv_desc) · line 2759
Quoted 10 lines
hw/i386/intel_iommu.c +9
-0+g_hash_table_foreach_remove (s->vtd_pasid_as, + vtd_flush_pasid, &pc_info); + s->pasid_cache_gen =1 ; +} +static int vtd_pasid_cache_gsi (IntelIOMMUState *s) { +trace_vtd_pasid_cache_gsi (); + +vtd_iommu_lock (s); +vtd_pasid_cache_reset (s);[1]
Quoted 12 lines
hw/i386/intel_iommu.c +9
-0+ +/* +* TODO: Global PASID cache invalidation may be +* flushes all the pasid caches. To be safe, after +* invalidating the pasid caches, emulator needs +* to replay the pasid bindings by walking guest +* pasid dir and pasid table. +*/ +vtd_iommu_unlock (s);return 0 ; }
[...]
Quoted 79 lines
Quoted 72 lines
Quoted 63 lines
Quoted 29 lines
(quoted diff) +0
-0
b/include/hw/i386/intel_iommu.hindex 4158116..3cc4b74 100644include/hw/i386/intel_iommu.h +0
-0@@ -69,6 +69,8 @@ typedef union VTD_IR_MSIAddress
VTD_IR_MSIAddress; typedef struct VTDPASIDDirEntry VTDPASIDDirEntry; typedef struct VTDPASIDEntry VTDPASIDEntry; typedef struct VTDIOMMUContext VTDIOMMUContext;include/hw/i386/intel_iommu.h +14
-0+typedef struct VTDPASIDCacheEntry VTDPASIDCacheEntry ;typedef +struct VTDPASIDAddressSpace VTDPASIDAddressSpace ;/* Context-Entry */ struct VTDContextEntry {@@ -101,6 +103,31 @@ struct VTDPASIDEntry {uint64_t val[8 ]; }; +struct pasid_key { +uint32_t pasid; +uint16_t sid; +}; + +struct VTDPASIDCacheEntry { +/* +* The cache entry is obsolete if +* pasid_cache_gen!=IntelIOMMUState.pasid_cache_gen +*/ +uint32_t pasid_cache_gen; +struct VTDPASIDEntry pasid_entry; };include/hw/i386/intel_iommu.h struct VTDPASIDEntry { · line 127
Quoted 10 lines
include/hw/i386/intel_iommu.h +10
-0+ +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;In
vtd_pasid_cache_gsi()[1], you directly resetpasid_cache_genfor each pasid address space. You never increase pasid_cache_entry.pasid_cache_gen. Then IIUC the gen will always be either 0 or 1. And...Quoted 13 lines
include/hw/i386/intel_iommu.h +3
-0+}; +struct VTDAddressSpace {PCIBus *bus;uint8_t devfn;@@ -122,6 +149,7 @@ struct VTDIOMMUContext {uint8_t devfn;IOMMUContext iommu_context;DualStageIOMMUObject *dsi_obj; +QLIST_ENTRY (VTDIOMMUContext) next;IntelIOMMUState *iommu_state; };include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 301
Quoted 2 lines
include/hw/i386/intel_iommu.h +0
-0@@ -272,9 +300,14 @@ struct IntelIOMMUState {GHashTable *vtd_as_by_busptr;/* VTDBus objects indexed by PCIBus* reference */
include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 302
Quoted 1 line
include/hw/i386/intel_iommu.h +0
-0VTDBus *vtd_as_by_bus_num[VTD_PCI_BUS_MAX ];/* VTDBus objects indexed by bus number */
Quoted 8 lines
include/hw/i386/intel_iommu.h struct IntelIOMMUState { · line 304
Quoted 2 lines
include/hw/i386/intel_iommu.h +2
-0+GHashTable *vtd_pasid_as;/* VTDPASIDAddressSpace instances */ +uint32_t pasid_cache_gen;/* Should be in [1,MAX] */ ... This should always be 1.
IIUC you can drop both of thepasid_cache_gen, because in this whole patchset you'll remove the pasid hash entry when it is invalidated, right? Then if the hash entry is there, it must be valid. When it's out-dated, it'll be removed from the hash.Quoted 4 lines
Oh, yes it is. However, it's not my intetion. I'd like to let [1] to increase the
s->pasid_cache_geninstead of justing zero it. I think it will save some time as loop hash table takes time. Thanks for catching it. :-)OK that's fine too. Then remember to conditionally reset it:
Quoted 12 lines
static int vtd_pasid_cache_gsi(IntelIOMMUState *s) {
trace_vtd_pasid_cache_gsi();
vtd_iommu_lock(s);
s->pasid_cache_gen++;
if (s->pasid_cache_gen >= THRESHOLD) {
vtd_pasid_cache_reset(s);
}
vtd_iommu_unlock(s);
return 0;
}thanks for the pseudo code. :-)
Regards,
Yi Liu
Re: [RFC v3 16/25] intel_iommu: add PASID cache management infrastructure
Quoted 4 lines · Yi L
Quoted 4 lines
- Remove the
vtd_pasid_ascheck right below because it's not needed.Quoted 2 lines
Quoted 8 lines
Quoted 2 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
+ if (
vtd_pasid_as&&^^^^^^^^^^^^yes, it is. In current series
vtd_add_find_pasid_as()doesn’t check the result ofvtd_pasid_asmem allocation, so no need to checkvtd_pasid_ashere either. However, it might be better to check the allocation result or it will result in issue if allocation failed. What's your preference here?
That should not be needed, because IIRC g_malloc0() will directly coredump if allocation fails. Even if not, it'll coredump in vtd_add_find_pasid_as() soon when accessing the NULL pointer.
Signature · 1 line
Peter Xu
RE: [RFC v3 16/25] intel_iommu: add PASID cache management infrastructure
Quoted 23 lines
-----Original Message-----
From: Peter Xu <[email protected]>
Sent: Thursday, February 13, 2020 11:14 PM
To: Liu, Yi L <[email protected]>
Subject: Re: [RFC v3 16/25]intel_iommu: add PASID cache management infrastructureQuoted 4 lines · Yi L
Quoted 4 lines
- Remove the
vtd_pasid_ascheck right below because it's not needed.Quoted 2 lines
Quoted 8 lines
Quoted 2 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
+ if (
vtd_pasid_as&&^^^^^^^^^^^^yes, it is. In current series
vtd_add_find_pasid_as()doesn’t check the result ofvtd_pasid_asmem allocation, so no need to checkvtd_pasid_ashere either. However, it might be better to check the allocation result or it will result in issue if allocation failed. What's your preference here?
Quoted 3 lines
That should not be needed, because IIRC
g_malloc0()will directly coredump if allocation fails. Even if not, it'll coredump invtd_add_find_pasid_as()soon when accessing the NULL pointer.
Cool, thanks for this message. Then I'll follow your suggestion to remove the vtd_pasid_as check.
Regards,
Yi Liu
RFC v3 17/25 vfio: add bind stage-1 page table support
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This patch adds bind_stage1_pgtbl() definition in DualStageIOMMUOops, 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/dual_stage_iommu.c | 26 ++++++++++++++++++++ hw/vfio/common.c | 48 +++++++++++++++++++++++++++++++++++++ include/hw/iommu/dual_stage_iommu.h | 22 +++++++++++++++++ 3 files changed, 96 insertions(+)
hw/iommu/dual_stage_iommu.c +26 -0
@@ -47,6 +47,32 @@ int ds_iommu_pasid_free(DualStageIOMMUObject *dsi_obj, uint32_t pasid)return -ENOENT ; } +int ds_iommu_bind_stage1_pgtbl (DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1BindData *data) +{ +if (!dsi_obj) { +return -ENOENT ; + } + +if (dsi_obj->ops && dsi_obj->ops->bind_stage1_pgtbl) { +return dsi_obj->ops->bind_stage1_pgtbl (dsi_obj, data); + } +return -ENOENT ; +} + +int ds_iommu_unbind_stage1_pgtbl (DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1BindData *data) +{ +if (!dsi_obj) { +return -ENOENT ; + } + +if (dsi_obj->ops && dsi_obj->ops->unbind_stage1_pgtbl) { +return dsi_obj->ops->unbind_stage1_pgtbl (dsi_obj, data); + } +return -ENOENT ; +} +void ds_iommu_object_init (DualStageIOMMUObject *dsi_obj,DualStageIOMMUOps *ops,DualStageIOMMUInfo *uinfo)
hw/vfio/common.c +48 -0
57 lines not on this page · raw message
include/hw/iommu/dual_stage_iommu.h +22 -0
@@ -31,6 +31,7 @@typedef struct DualStageIOMMUObject DualStageIOMMUObject ;typedef struct DualStageIOMMUOps DualStageIOMMUOps ;typedef struct DualStageIOMMUInfo DualStageIOMMUInfo ; +typedef struct DualIOMMUStage1BindData DualIOMMUStage1BindData ;struct DualStageIOMMUOps {/* Allocate pasid from DualStageIOMMU (a.k.a. host IOMMU) */
@@ -41,6 +42,16 @@ struct DualStageIOMMUOps {/* Reclaim a pasid from DualStageIOMMU (a.k.a. host IOMMU) */ int (*pasid_free)(DualStageIOMMUObject *dsi_obj,uint32_t pasid); +/* +* Bind stage-1 page table to a DualStageIOMMU (a.k.a. host +* IOMMU which has dual stage DMA translation capability. +* @bind_data specifies the bind configurations. +*/ +int (*bind_stage1_pgtbl)(DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1BindData *bind_data); +/* Undo a previous bind. @bind_data specifies the unbind info. */ +int (*unbind_stage1_pgtbl)(DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1BindData *bind_data); };struct DualStageIOMMUInfo {
@@ -55,9 +66,20 @@ struct DualStageIOMMUObject {DualStageIOMMUInfo uinfo; }; +struct DualIOMMUStage1BindData { +uint32_t pasid; +union { +struct iommu_gpasid_bind_data gpasid_bind; + } bind_data; +}; +int ds_iommu_pasid_alloc (DualStageIOMMUObject *dsi_obj,uint32_t min,uint32_t max,uint32_t *pasid);int ds_iommu_pasid_free (DualStageIOMMUObject *dsi_obj,uint32_t pasid); +int ds_iommu_bind_stage1_pgtbl (DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1BindData *bind_data); +int ds_iommu_unbind_stage1_pgtbl (DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1BindData *bind_data); void ds_iommu_object_init(DualStageIOMMUObject *dsi_obj, DualStageIOMMUOps *ops,
Signature · 1 line
2.7.4
RFC v3 18/25 intel_iommu: bind/unbind guest page table to host
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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 a 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 | 83 +++++++++++++++++++++++++++++++++++++++++- hw/i386/intel_iommu_internal.h | 26 +++++++++++++ 2 files changed, 107 insertions(+), 2 deletions(-)
hw/i386/intel_iommu.c +81 -2
@@ -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) \
@@ -695,6 +696,16 @@ 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_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 ;
@@ -1854,6 +1865,68 @@ static void vtd_context_global_invalidate(IntelIOMMUState *s)vtd_iommu_replay_all (s); } +static int vtd_bind_guest_pasid (IntelIOMMUState *s,VTDBus *vtd_bus, +int devfn,int pasid,VTDPASIDEntry *pe,VTDPASIDOp op) +{ +VTDIOMMUContext *vtd_icx; +DualIOMMUStage1BindData bind_data; +struct iommu_gpasid_bind_data *g_bind_data; +int ret =-1 ; + + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +return ret; + } + +if (vtd_icx->dsi_obj->uinfo.pasid_format + !=IOMMU_PASID_FORMAT_INTEL_VTD ) + { +error_report_once ("Dual Stage IOMMU is not compatible!!\n" ); + } + + bind_data.pasid = pasid; + g_bind_data = &bind_data.bind_data.gpasid_bind; + + g_bind_data->flags =0 ; + g_bind_data->vtd.flags =0 ; +switch (op) { +case VTD_PASID_BIND : +case VTD_PASID_UPDATE : + g_bind_data->version =IOMMU_UAPI_VERSION ; + 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->vtd.flags = + (VTD_SM_PASID_ENTRY_SRE_BIT (pe->val[2 ]) ?1 :0 ) + | (VTD_SM_PASID_ENTRY_EAFE_BIT (pe->val[2 ]) ?1 :0 ) + | (VTD_SM_PASID_ENTRY_PCD_BIT (pe->val[1 ]) ?1 :0 ) + | (VTD_SM_PASID_ENTRY_PWT_BIT (pe->val[1 ]) ?1 :0 ) + | (VTD_SM_PASID_ENTRY_EMTE_BIT (pe->val[1 ]) ?1 :0 ) + | (VTD_SM_PASID_ENTRY_CD_BIT (pe->val[1 ]) ?1 :0 ); + g_bind_data->vtd.pat =VTD_SM_PASID_ENTRY_PAT (pe->val[1 ]); + g_bind_data->vtd.emt =VTD_SM_PASID_ENTRY_EMT (pe->val[1 ]); + ret =ds_iommu_bind_stage1_pgtbl (vtd_icx->dsi_obj, &bind_data); +break ; +case VTD_PASID_UNBIND : + g_bind_data->version =IOMMU_UAPI_VERSION ; + g_bind_data->format =IOMMU_PASID_FORMAT_INTEL_VTD ; + g_bind_data->gpgd =0 ; + g_bind_data->addr_width =0 ; + g_bind_data->hpasid = pasid; + g_bind_data->gpasid = pasid; + g_bind_data->flags |=IOMMU_SVA_GPASID_VAL ; + ret =ds_iommu_unbind_stage1_pgtbl (vtd_icx->dsi_obj, &bind_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 */
@@ -2532,18 +2605,20 @@ static gboolean vtd_flush_pasid(gpointer key, gpointer value,/* pasid entry was updated, thus update the pasid cache */ pc_entry->pasid_entry = pe; pc_entry->pasid_cache_gen = s->pasid_cache_gen; +vtd_bind_guest_pasid (s, vtd_bus, devfn, + pasid, &pe,VTD_PASID_UPDATE );/* * 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 false; remove: +vtd_bind_guest_pasid (s, vtd_bus, devfn, + pasid,NULL ,VTD_PASID_UNBIND );/* * TODO: -* - send pasid unbind to host for passthru devices * - when pasid-base-iotlb(piotlb) infrastructure is ready, * should invalidate QEMU piotlb togehter with this change. */
@@ -2634,6 +2709,10 @@ static inline void vtd_fill_in_pe_cache( pc_entry->pasid_entry = *pe; pc_entry->pasid_cache_gen = s->pasid_cache_gen; +vtd_bind_guest_pasid (s, vtd_pasid_as->vtd_bus, + vtd_pasid_as->devfn, + vtd_pasid_as->pasid, + pe,VTD_PASID_BIND ); } static int vtd_pasid_cache_psi(IntelIOMMUState *s,
hw/i386/intel_iommu_internal.h +26 -0
@@ -486,6 +486,20 @@ struct VTDRootEntry { };typedef struct VTDRootEntry VTDRootEntry ; +enum VTD_DUAL_STAGE_UAPI { +UAPI_BIND_GPASID , +UAPI_NUM +}; +typedef enum VTD_DUAL_STAGE_UAPI VTD_DUAL_STAGE_UAPI ; + +enum VTDPASIDOp { +VTD_PASID_BIND , +VTD_PASID_UNBIND , +VTD_PASID_UPDATE , +VTD_OP_NUM +}; +typedef enum VTDPASIDOp VTDPASIDOp ; +struct VTDPASIDCacheInfo {#define VTD_PASID_CACHE_GLOBAL (1ULL << 0)#define VTD_PASID_CACHE_DOMSI (1ULL << 1)
@@ -556,6 +570,18 @@ 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) +/* Adjusted guest-address-width */ +#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 v3 19/25 intel_iommu: replay guest pasid bindings to host
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This patch adds guest pasid bindings replay for domain selective pasid cache invalidation(dsi) and global pasid cache invalidation by walking guest pasid table.
Reason:
Guest OS may flush the pasid cache with a larger granularity. e.g. guest does a svm_bind() but flush the pasid cache with global or domain selective pasid cache invalidation instead of pasid selective(psi) pasid cache invalidation. Regards to such case, it works in host. Per spec, a global or domain selective pasid cache invalidation should be able to cover what a pasid selective invalidation does. The only concern is performance deduction since dsi and global cache invalidation will flush more than psi. To align with native, vIOMMU needs emulator needs to do replay for the two invalidation granularity to reflect the latest pasid bindings in guest pasid table.
--- hw/i386/intel_iommu.c | 156 ++++++++++++++++++++++++++++++++++++++--- hw/i386/intel_iommu_internal.h | 1 + 2 files changed, 147 insertions(+), 10 deletions(-)
hw/i386/intel_iommu.c +146 -10
@@ -68,6 +68,8 @@ 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 int vtd_update_pe_cache_for_dev (IntelIOMMUState *s, +VTDBus *vtd_bus,int devfn,int pasid,VTDPASIDEntry *pe);static void vtd_panic_require_caching_mode (void ) {
@@ -2625,6 +2627,113 @@ remove:return true; } +/** +* Constant information used during pasid table walk +* @vtd_icx: VTDIOMMUContext +* @flags: indicates if it is domain selective walk +* @did: domain ID of the pasid table walk +*/ +typedef struct { +VTDIOMMUContext *vtd_icx; +#define VTD_PASID_TABLE_DID_SEL_WALK (1ULL << 0); +uint32_t flags; +uint16_t did; +}vtd_pasid_table_walk_info ; + +static bool vtd_sm_pasid_table_walk_one (IntelIOMMUState *s, +dma_addr_t pt_base, +int start, +int end, +vtd_pasid_table_walk_info *info) +{ +VTDPASIDEntry pe; +int pasid = start; +int pasid_next; +VTDIOMMUContext *vtd_icx = info->vtd_icx; + +while (pasid < end) { + pasid_next = pasid +1 ; + +if (!vtd_get_pe_in_pasid_leaf_table (s, pasid, pt_base, &pe) + &&vtd_pe_present (&pe)) { +if (vtd_update_pe_cache_for_dev (s, vtd_icx->vtd_bus, + vtd_icx->devfn, pasid, &pe)) { +error_report_once ("%s, bus: %d, devfn: %d, pasid: %d" , + __func__, +pci_bus_num (vtd_icx->vtd_bus->bus), + vtd_icx->devfn, pasid); +return false; + } + } + pasid = pasid_next; + } +return true; +} + +/* +* 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. +*/ +static void vtd_sm_pasid_table_walk (IntelIOMMUState *s,dma_addr_t pdt_base, +int start,int end,vtd_pasid_table_walk_info *info) +{ +VTDPASIDDirEntry pdire; +int pasid = start; +int pasid_next; +dma_addr_t pt_base; + +while (pasid < end) { + pasid_next = pasid +VTD_PASID_TBL_ENTRY_NUM ; +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 ; +if (!vtd_sm_pasid_table_walk_one (s, + pt_base, pasid, pasid_next, info)) { +break ; + } + } + pasid = pasid_next; + } +} + +/** +* This function replay the guest pasid bindings to hots by +* walking the guest PASID table. This ensures host will have +* latest guest pasid bindings. +*/ +static void vtd_replay_guest_pasid_bindings (IntelIOMMUState *s, +uint16_t *did,bool is_dsi) +{ +VTDContextEntry ce; +vtd_pasid_table_walk_info info; +VTDIOMMUContext *vtd_icx; + +if (is_dsi) { + info.flags =VTD_PASID_TABLE_DID_SEL_WALK ; + info.did = *did; + } + +/* +* In this replay, only needs to care about the devices which +* has iommu_context created. For the one not have iommu_context, +* it is not necessary to replay the bindings since their cache +* could be re-created in the next DMA address transaltion. +*/ + QLIST_FOREACH(vtd_icx, &s->vtd_dev_icx_list, next) { +if (!vtd_dev_to_context_entry (s, +pci_bus_num (vtd_icx->vtd_bus->bus), + vtd_icx->devfn, &ce)) { + info.vtd_icx = vtd_icx; +vtd_sm_pasid_table_walk (s, +VTD_CE_GET_PASID_DIR_TABLE (&ce), +0 , +VTD_MAX_HPASID , + &info); + } + } +} +static int vtd_pasid_cache_dsi (IntelIOMMUState *s,uint16_t domain_id) {VTDPASIDCacheInfo pc_info;
@@ -2642,12 +2751,13 @@ static int vtd_pasid_cache_dsi(IntelIOMMUState *s, uint16_t domain_id) vtd_flush_pasid, &pc_info);/* -* TODO: Domain selective PASID cache invalidation -* flushes all the pasid caches within a domain. To -* be safe, after invalidating the pasid caches, emulator -* needs to replay the pasid bindings by walking guest -* pasid dir and pasid table. +* Domain selective PASID cache invalidation flushes +* all the pasid caches within a domain. To be safe, +* after invalidating the pasid caches, emulator needs +* to replay the pasid bindings by walking guest pasid +* dir and pasid table. */ +vtd_replay_guest_pasid_bindings (s, &domain_id, true);vtd_iommu_unlock (s);return 0 ; }
@@ -2715,6 +2825,31 @@ static inline void vtd_fill_in_pe_cache( pe, VTD_PASID_BIND); } +/** +* This function updates the pasid entry cached in &vtd_pasid_as. +* Caller of this function should hold iommu_lock. +*/ +static int vtd_update_pe_cache_for_dev (IntelIOMMUState *s,VTDBus *vtd_bus, +int devfn,int pasid,VTDPASIDEntry *pe) +{ +VTDPASIDAddressSpace *vtd_pasid_as; + + vtd_pasid_as =vtd_add_find_pasid_as (s, vtd_bus, + devfn, pasid, true); +if (!vtd_pasid_as) { +error_report_once ("%s, fatal error happened!\n" , __func__); +return -1 ; + } + +if (vtd_pasid_as->pasid_cache_entry.pasid_cache_gen == + s->pasid_cache_gen) { +return 0 ; + } + +vtd_fill_in_pe_cache (vtd_pasid_as, pe); +return 0 ; +} +static int vtd_pasid_cache_psi (IntelIOMMUState *s,uint16_t domain_id,uint32_t pasid) {
@@ -2838,12 +2973,13 @@ static int vtd_pasid_cache_gsi(IntelIOMMUState *s)vtd_pasid_cache_reset (s);/* -* TODO: Global PASID cache invalidation may be -* flushes all the pasid caches. To be safe, after -* invalidating the pasid caches, emulator needs -* to replay the pasid bindings by walking guest -* pasid dir and pasid table. +* Global PASID cache invalidation flushes all +* the pasid caches. To be safe, after invalidating +* the pasid caches, emulator needs to replay the +* pasid bindings by walking guest pasid dir and +* pasid table. */ +vtd_replay_guest_pasid_bindings (s,NULL , false);vtd_iommu_unlock (s);return 0 ; }
hw/i386/intel_iommu_internal.h +1 -0
@@ -558,6 +558,7 @@ typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo;#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
Signature · 1 line
2.7.4
RFC v3 20/25 intel_iommu: replay pasid binds after context cache invalidation
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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 | 67 ++++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 6 +++- hw/i386/trace-events | 1 + 3 files changed, 73 insertions(+), 1 deletion(-)
hw/i386/intel_iommu.c +67 -0
@@ -70,6 +70,10 @@ static void vtd_address_space_unmap(VTDAddressSpace *as, IOMMUNotifier *n);static void vtd_pasid_cache_reset (IntelIOMMUState *s);static int vtd_update_pe_cache_for_dev (IntelIOMMUState *s,VTDBus *vtd_bus,int devfn,int pasid,VTDPASIDEntry *pe); +static void vtd_replay_guest_pasid_bindings (IntelIOMMUState *s, +uint16_t *did,bool is_dsi); +static void vtd_pasid_cache_devsi (IntelIOMMUState *s, +VTDBus *vtd_bus,uint16_t devfn);static void vtd_panic_require_caching_mode (void ) {
@@ -1865,6 +1869,10 @@ static void vtd_context_global_invalidate(IntelIOMMUState *s)* VT-d emulation codes. */ vtd_iommu_replay_all (s); + +vtd_iommu_lock (s); +vtd_replay_guest_pasid_bindings (s,NULL , false); +vtd_iommu_unlock (s); } static int vtd_bind_guest_pasid(IntelIOMMUState *s, VTDBus *vtd_bus,
@@ -1986,6 +1994,22 @@ 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) +* modify the pasid cache gen and pasid-based iotlb gen +* value (will be added in following patches) +* 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); } } }
@@ -2581,6 +2605,12 @@ static gboolean vtd_flush_pasid(gpointer key, gpointer value,/* Fall through */ case VTD_PASID_CACHE_GLOBAL :break ; +case VTD_PASID_CACHE_DEVSI : +if (pc_info->vtd_bus != vtd_bus || + pc_info->devfn == devfn) { +return false; + } +break ;default :return false; }
@@ -2944,6 +2974,43 @@ static int vtd_pasid_cache_psi(IntelIOMMUState *s,return 0 ; } +static void vtd_pasid_cache_devsi (IntelIOMMUState *s, +VTDBus *vtd_bus,uint16_t devfn) +{ +VTDPASIDCacheInfo pc_info; +VTDContextEntry ce; +vtd_pasid_table_walk_info info; + +trace_vtd_pasid_cache_devsi (devfn); + + pc_info.flags =VTD_PASID_CACHE_DEVSI ; + pc_info.vtd_bus = vtd_bus; + pc_info.devfn = devfn; + +vtd_iommu_lock (s); +g_hash_table_foreach_remove (s->vtd_pasid_as, vtd_flush_pasid, &pc_info); + +/* +* To be safe, after invalidating the pasid caches, +* emulator needs to replay the pasid bindings by +* walking guest pasid dir and pasid table. +*/ +if (vtd_bus->dev_icx[devfn] && + !vtd_dev_to_context_entry (s, +pci_bus_num (vtd_bus->bus), + devfn, &ce)) { + info.flags =0x0 ; + info.did =0 ; + info.vtd_icx = vtd_bus->dev_icx[devfn]; +vtd_sm_pasid_table_walk (s, +VTD_CE_GET_PASID_DIR_TABLE (&ce), +0 , +VTD_MAX_HPASID , + &info); + } +vtd_iommu_unlock (s); +} +/** * Caller of this function should hold iommu_lock */
hw/i386/intel_iommu_internal.h +5 -1
@@ -504,13 +504,17 @@ struct VTDPASIDCacheInfo {#define VTD_PASID_CACHE_GLOBAL (1ULL << 0)#define VTD_PASID_CACHE_DOMSI (1ULL << 1)#define VTD_PASID_CACHE_PASIDSI (1ULL << 2) +#define VTD_PASID_CACHE_DEVSI (1ULL << 3)uint32_t flags;uint16_t domain_id;uint32_t pasid; +VTDBus *vtd_bus; +uint16_t devfn; };#define VTD_PASID_CACHE_INFO_MASK (VTD_PASID_CACHE_GLOBAL | \ VTD_PASID_CACHE_DOMSI | \ - VTD_PASID_CACHE_PASIDSI) + VTD_PASID_CACHE_PASIDSI | \ + VTD_PASID_CACHE_DEVSI)typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ;/* Masks for struct VTDRootEntry */
hw/i386/trace-events +1 -0
@@ -26,6 +26,7 @@ vtd_pasid_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_pasid_cache_devsi(uint16_t devfn) "Dev slective 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 v3 21/25 intel_iommu: do not pass down pasid bind for PASID #0
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
RID_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://lkml.org/lkml/2019/9/23/297
--- hw/i386/intel_iommu.c | 10 ++++++++++ 1 file changed, 10 insertions(+)
hw/i386/intel_iommu.c +10 -0
@@ -1883,6 +1883,16 @@ static int vtd_bind_guest_pasid(IntelIOMMUState *s, VTDBus *vtd_bus,struct iommu_gpasid_bind_data *g_bind_data;int ret =-1 ; +if (pasid <VTD_MIN_HPASID ) { +/* +* 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_icx = vtd_bus->dev_icx[devfn];if (!vtd_icx) {return ret;
Signature · 1 line
2.7.4
RFC v3 22/25 vfio: add support for flush iommu stage-1 cache
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This patch adds flush_stage1_cache() definition in DualStageIOMMUOops. 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.
--- hw/iommu/dual_stage_iommu.c | 13 +++++++++++++ hw/vfio/common.c | 24 ++++++++++++++++++++++++ include/hw/iommu/dual_stage_iommu.h | 14 ++++++++++++++ 3 files changed, 51 insertions(+)
hw/iommu/dual_stage_iommu.c +13 -0
@@ -73,6 +73,19 @@ int ds_iommu_unbind_stage1_pgtbl(DualStageIOMMUObject *dsi_obj,return -ENOENT ; } +int ds_iommu_flush_stage1_cache (DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1Cache *cache) +{ +if (!dsi_obj) { +return -ENOENT ; + } + +if (dsi_obj->ops && dsi_obj->ops->flush_stage1_cache) { +return dsi_obj->ops->flush_stage1_cache (dsi_obj, cache); + } +return -ENOENT ; +} +void ds_iommu_object_init (DualStageIOMMUObject *dsi_obj,DualStageIOMMUOps *ops,DualStageIOMMUInfo *uinfo)
hw/vfio/common.c +24 -0
@@ -1265,11 +1265,35 @@ static int vfio_ds_iommu_unbind_stage1_pgtbl(DualStageIOMMUObject *dsi_obj,return ret; } +static int vfio_ds_iommu_flush_stage1_cache (DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1Cache *cache) +{ +VFIOContainer *container =container_of (dsi_obj, VFIOContainer, dsi_obj); +struct vfio_iommu_type1_cache_invalidate *cache_inv; +unsigned long argsz; +int ret =0 ; + + argsz =sizeof (*cache_inv) +sizeof (cache->cache_info); + cache_inv =g_malloc0 (argsz); + cache_inv->argsz = argsz; + cache_inv->flags =0 ; +memcpy (&cache_inv->cache_info, &cache->cache_info, +sizeof (cache->cache_info)); + +if (ioctl (container->fd,VFIO_IOMMU_CACHE_INVALIDATE , cache_inv)) { +error_report ("%s: iommu cache flush failed: %d" , __func__, -errno); + ret = -errno; + } +g_free (cache_inv); +return ret; +} +static struct DualStageIOMMUOps vfio_ds_iommu_ops = { .pasid_alloc = vfio_ds_iommu_pasid_alloc, .pasid_free = vfio_ds_iommu_pasid_free, .bind_stage1_pgtbl = vfio_ds_iommu_bind_stage1_pgtbl, .unbind_stage1_pgtbl = vfio_ds_iommu_unbind_stage1_pgtbl, + .flush_stage1_cache = vfio_ds_iommu_flush_stage1_cache, }; static int vfio_get_iommu_info(VFIOContainer *container,
include/hw/iommu/dual_stage_iommu.h +14 -0
@@ -32,6 +32,7 @@ typedef struct DualStageIOMMUObject DualStageIOMMUObject;typedef struct DualStageIOMMUOps DualStageIOMMUOps ;typedef struct DualStageIOMMUInfo DualStageIOMMUInfo ;typedef struct DualIOMMUStage1BindData DualIOMMUStage1BindData ; +typedef struct DualIOMMUStage1Cache DualIOMMUStage1Cache ;struct DualStageIOMMUOps {/* Allocate pasid from DualStageIOMMU (a.k.a. host IOMMU) */
@@ -52,6 +53,12 @@ struct DualStageIOMMUOps {/* Undo a previous bind. @bind_data specifies the unbind info. */ int (*unbind_stage1_pgtbl)(DualStageIOMMUObject *dsi_obj,DualIOMMUStage1BindData *bind_data); +/* +* Propagate stage-1 cache flush to DualStageIOMMU (a.k.a. +* host IOMMU), cache info specifid in @cache +*/ +int (*flush_stage1_cache)(DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1Cache *cache); };struct DualStageIOMMUInfo {
@@ -73,6 +80,11 @@ struct DualIOMMUStage1BindData { } bind_data; }; +struct DualIOMMUStage1Cache { +uint32_t pasid; +struct iommu_cache_invalidate_info cache_info; +}; +int ds_iommu_pasid_alloc (DualStageIOMMUObject *dsi_obj,uint32_t min,uint32_t max,uint32_t *pasid);int ds_iommu_pasid_free (DualStageIOMMUObject *dsi_obj,uint32_t pasid);
@@ -80,6 +92,8 @@ int ds_iommu_bind_stage1_pgtbl(DualStageIOMMUObject *dsi_obj, DualIOMMUStage1BindData *bind_data);int ds_iommu_unbind_stage1_pgtbl (DualStageIOMMUObject *dsi_obj,DualIOMMUStage1BindData *bind_data); +int ds_iommu_flush_stage1_cache (DualStageIOMMUObject *dsi_obj, +DualIOMMUStage1Cache *cache); void ds_iommu_object_init(DualStageIOMMUObject *dsi_obj, DualStageIOMMUOps *ops,
Signature · 1 line
2.7.4
RFC v3 23/25 intel_iommu: process PASID-based iotlb invalidation
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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 | 57 ++++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 13 ++++++++++ 2 files changed, 70 insertions(+)
hw/i386/intel_iommu.c +57 -0
@@ -3102,6 +3102,59 @@ static bool vtd_process_pasid_desc(IntelIOMMUState *s,return (ret ==0 ) ? true : false; } +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 ]); +if (am >VTD_MAMV ) { +error_report_once ("Invalid am, > max am value, hi: 0x%" PRIx64 +" lo: 0x%" PRIx64, inv_desc->val[1 ], inv_desc->val[0 ]); +return false; + } +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) {
@@ -3216,6 +3269,10 @@ 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
@@ -461,6 +461,19 @@ 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 v3 24/25 intel_iommu: propagate PASID-based iotlb invalidation to host
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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.
--- hw/i386/intel_iommu.c | 122 +++++++++++++++++++++++++++++++++++++++++ hw/i386/intel_iommu_internal.h | 7 +++ 2 files changed, 129 insertions(+)
hw/i386/intel_iommu.c +122 -0
@@ -3102,15 +3102,137 @@ static bool vtd_process_pasid_desc(IntelIOMMUState *s,return (ret ==0 ) ? true : false; } +static void vtd_invalidate_piotlb (IntelIOMMUState *s,VTDBus *vtd_bus, +int devfn,DualIOMMUStage1Cache *stage1_cache) +{ +VTDIOMMUContext *vtd_icx; + vtd_icx = vtd_bus->dev_icx[devfn]; +if (!vtd_icx) { +return ; + } +if (ds_iommu_flush_stage1_cache (vtd_icx->dsi_obj, stage1_cache)) { +error_report ("Cache flush failed" ); + } +} + +static inline bool vtd_pasid_cache_valid ( +VTDPASIDAddressSpace *vtd_pasid_as) +{ +return (vtd_pasid_as->iommu_state->pasid_cache_gen && + (vtd_pasid_as->iommu_state->pasid_cache_gen + == vtd_pasid_as->pasid_cache_entry.pasid_cache_gen)); +} + +/** +* 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; +uint16_t did; + +/* +* Needs to check whether the pasid entry cache stored in +* vtd_pasid_as is valid or not. "invalid" means the pasid +* cache has been flushed, thus host should have done piotlb +* invalidation together with a pasid cache invalidation, so +* no need to pass down piotlb invalidation to host for better +* performance. Only when pasid entry cache is "valid", should +* a piotlb invalidation be propagated to host since it means +* guest just modified a mapping in its page table. +*/ +if (!vtd_pasid_cache_valid (vtd_pasid_as)) { +return ; + } + + did =vtd_pe_get_domain_id ( + &(vtd_pasid_as->pasid_cache_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->stage1_cache); + } + +/* +* 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; +DualIOMMUStage1Cache stage1_cache; + + stage1_cache.pasid = pasid; + + cache_info = &stage1_cache.cache_info; + cache_info->version =IOMMU_UAPI_VERSION ; + cache_info->cache =IOMMU_CACHE_INV_TYPE_IOTLB ; + cache_info->granularity =IOMMU_INV_GRANU_PASID ; + cache_info->pasid_info.pasid = pasid; + cache_info->pasid_info.flags =IOMMU_INV_PASID_FLAGS_PASID ; + + piotlb_info.domain_id = domain_id; + piotlb_info.pasid = pasid; + piotlb_info.stage1_cache = &stage1_cache; + +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); }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; +DualIOMMUStage1Cache stage1_cache; + + stage1_cache.pasid = pasid; + + cache_info = &stage1_cache.cache_info; + cache_info->version =IOMMU_UAPI_VERSION ; + cache_info->cache =IOMMU_CACHE_INV_TYPE_IOTLB ; + cache_info->granularity =IOMMU_INV_GRANU_ADDR ; + cache_info->addr_info.flags =IOMMU_INV_ADDR_FLAGS_PASID ; + cache_info->addr_info.flags |= ih ?IOMMU_INV_ADDR_FLAGS_LEAF :0 ; + cache_info->addr_info.pasid = pasid; + cache_info->addr_info.addr = addr; + cache_info->addr_info.granule_size =1 << (12 + am); + cache_info->addr_info.nb_granules =1 ; + + piotlb_info.domain_id = domain_id; + piotlb_info.pasid = pasid; + piotlb_info.stage1_cache = &stage1_cache; + +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); } static bool vtd_process_piotlb_desc(IntelIOMMUState *s,
hw/i386/intel_iommu_internal.h +7 -0
@@ -530,6 +530,13 @@ struct VTDPASIDCacheInfo { VTD_PASID_CACHE_DEVSI)typedef struct VTDPASIDCacheInfo VTDPASIDCacheInfo ; +struct VTDPIOTLBInvInfo { +uint16_t domain_id; +uint32_t pasid; +DualIOMMUStage1Cache *stage1_cache; +}; +typedef struct VTDPIOTLBInvInfo VTDPIOTLBInvInfo ; +/* Masks for struct VTDRootEntry */ #define VTD_ROOT_ENTRY_P 1ULL#define VTD_ROOT_ENTRY_CTP (~0xfffULL)
Signature · 1 line
2.7.4
RFC v3 25/25 intel_iommu: process PASID-based Device-TLB invalidation
In reply to Liu, Yi LFrom: Liu Yi L <[email protected]>
This 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
@@ -3290,6 +3290,17 @@ 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) {
@@ -3411,6 +3422,13 @@ 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
@@ -390,6 +390,7 @@ 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
Patchew URL: https://patchew.org/QEMU/[email protected]/
Patchew URL: https://patchew.org/QEMU/[email protected]/
Hi,
This series failed the docker-mingw@fedora build test. Please find the testing commands and their output below. If you have Docker installed, you can probably reproduce it locally.
=== TEST SCRIPT BEGIN ===
#! /bin/bash
export ARCH=x86_64
make docker-image-fedora V=1 NETWORK=1
time make docker-test-mingw@fedora J=14 NETWORK=1
=== TEST SCRIPT END ===
from /tmp/qemu-test/src/include/hw/pci/pci_bus.h:4,
from /tmp/qemu-test/src/include/hw/pci-host/i440fx.h:15,
from /tmp/qemu-test/src/stubs/pci-host-piix.c:2:
/tmp/qemu-test/src/include/hw/iommu/dual_stage_iommu.h:26:10: fatal error: linux/iommu.h: No such file or directory
#include <linux/iommu.h>
^~~~~~~~~~~~~~~compilation terminated.
make: *** [/tmp/qemu-test/src/rules.mak:69: stubs/pci-host-piix.o] Error 1
make: *** Waiting for unfinished jobs....
Traceback (most recent call last):
File "./tests/docker/docker.py", line 662, in <module>
---
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['sudo', '-n', 'docker', 'run', '--label', 'com.qemu.instance.uuid=8a4150439dbc4c62abdee4366960ac9a', '-u', '1003', '--security-opt', 'seccomp=unconfined', '--rm', '-e', 'TARGET_LIST=', '-e', 'EXTRA_CONFIGURE_OPTS=', '-e', 'V=', '-e', 'J=14', '-e', 'DEBUG=', '-e', 'SHOW_ENV=', '-e', 'CCACHE_DIR=/var/tmp/ccache', '-v', '/home/patchew2/.cache/qemu-docker-ccache:/var/tmp/ccache:z', '-v', '/var/tmp/patchew-tester-tmp-qdowmtdx/src/docker-src.2020-01-29-08.42.04.21064:/var/tmp/qemu:z,ro', 'qemu:fedora', '/var/tmp/qemu/run', 'test-mingw']' returned non-zero exit status 2. filter=--filter=label=com.qemu.instance.uuid=8a4150439dbc4c62abdee4366960ac9a
make[1]: *** [docker-run] Error 1
make[1]: Leaving directory `/var/tmp/patchew-tester-tmp-qdowmtdx/src'
make: *** [docker-run-test-mingw@fedora] Error 2
real 2m5.234s
user 0m7.662s
The full log is available at
http://patchew.org/logs/[email protected]/testing.docker-mingw@fedora/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to [email protected]
Patchew URL: https://patchew.org/QEMU/[email protected]/
Patchew URL: https://patchew.org/QEMU/[email protected]/
Hi,
This series failed the docker-quick@centos7 build test. Please find the testing commands and their output below. If you have Docker installed, you can probably reproduce it locally.
=== TEST SCRIPT BEGIN ===
#!/bin/bash
make docker-image-centos7 V=1 NETWORK=1
time make docker-test-quick@centos7 SHOW_ENV=1 J=14 NETWORK=1 === TEST SCRIPT END ===
CC hw/pci/pci_host.o
CC hw/pci/pcie.o/tmp/qemu-test/src/hw/pci-host/designware.c: In function 'designware_pcie_host_realize':/tmp/qemu-test/src/hw/pci-host/designware.c:693:5: error: incompatible type for argument 2 of 'pci_setup_iommu'
pci_setup_iommu(pci->bus, designware_iommu_ops, s);
^
In file included from /tmp/qemu-test/src/include/hw/pci/msi.h:24:0,
---
/tmp/qemu-test/src/include/hw/pci/pci.h:499:6: note: expected 'const struct PCIIOMMUOps *' but argument is of type 'PCIIOMMUOps'
void pci_setup_iommu(PCIBus *bus, const PCIIOMMUOps *iommu_ops, void *opaque);
^make: *** [hw/pci-host/designware.o] Error 1
make: *** Waiting for unfinished jobs....
rm tests/qemu-iotests/socket_scm_helper.o
Traceback (most recent call last):
---
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['sudo', '-n', 'docker', 'run', '--label', 'com.qemu.instance.uuid=08fccc258f5241b886ff89ccf43e8926', '-u', '1001', '--security-opt', 'seccomp=unconfined', '--rm', '-e', 'TARGET_LIST=', '-e', 'EXTRA_CONFIGURE_OPTS=', '-e', 'V=', '-e', 'J=14', '-e', 'DEBUG=', '-e', 'SHOW_ENV=1', '-e', 'CCACHE_DIR=/var/tmp/ccache', '-v', '/home/patchew/.cache/qemu-docker-ccache:/var/tmp/ccache:z', '-v', '/var/tmp/patchew-tester-tmp-xrnho7rm/src/docker-src.2020-01-29-08.46.10.29742:/var/tmp/qemu:z,ro', 'qemu:centos7', '/var/tmp/qemu/run', 'test-quick']' returned non-zero exit status 2. filter=--filter=label=com.qemu.instance.uuid=08fccc258f5241b886ff89ccf43e8926
make[1]: *** [docker-run] Error 1
make[1]: Leaving directory `/var/tmp/patchew-tester-tmp-xrnho7rm/src'
make: *** [docker-run-test-quick@centos7] Error 2
real 2m13.263s
user 0m8.269s
The full log is available at
http://patchew.org/logs/[email protected]/testing.docker-quick@centos7/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to [email protected]