Hi Eric, On 2021/4/27 3:16, Auger Eric wrote:
Hi Kunkun,On 4/15/21 4:03 AM, Kunkun Jiang wrote:Hi Eric, On 2021/4/14 16:05, Auger Eric wrote:Hi Kunkun, On 4/14/21 3:45 AM, Kunkun Jiang wrote:On 2021/4/13 20:57, Auger Eric wrote:Hi Kunkun, On 4/13/21 2:10 PM, Kunkun Jiang wrote:Hi Eric, On 2021/4/11 20:08, Eric Auger wrote:In nested mode, legacy vfio_iommu_map_notify cannot be used as there is no "caching" mode and we do not trap on map. On Intel, vfio_iommu_map_notify was used to DMA map the RAM through the host single stage. With nested mode, we need to setup the stage 2 and the stage 1 separately. This patch introduces a prereg_listener to setup the stage 2 mapping. The stage 1 mapping, owned by the guest, is passed to the host when the guest invalidates the stage 1 configuration, through a dedicated PCIPASIDOps callback. Guest IOTLB invalidations are cascaded downto the host through another IOMMU MR UNMAP notifier. Signed-off-by: Eric Auger <eric.au...@redhat.com> --- v7 -> v8: - properly handle new IOMMUTLBEntry fields and especially propagate DOMAIN and PASID based invalidations v6 -> v7: - remove PASID based invalidation v5 -> v6: - add error_report_err() - remove the abort in case of nested stage case v4 -> v5: - use VFIO_IOMMU_SET_PASID_TABLE - use PCIPASIDOps for config notification v3 -> v4: - use iommu_inv_pasid_info for ASID invalidation v2 -> v3: - use VFIO_IOMMU_ATTACH_PASID_TABLE - new user API - handle leaf v1 -> v2: - adapt to uapi changes - pass the asid - pass IOMMU_NOTIFIER_S1_CFG when initializing the config notifier --- hw/vfio/common.c | 139 +++++++++++++++++++++++++++++++++++++++++-- hw/vfio/pci.c | 21 +++++++ hw/vfio/trace-events | 2 + 3 files changed, 157 insertions(+), 5 deletions(-) diff --git a/hw/vfio/common.c b/hw/vfio/common.c index 0cd7ef2139..e369d451e7 100644 --- a/hw/vfio/common.c +++ b/hw/vfio/common.c @@ -595,6 +595,73 @@ static bool vfio_get_xlat_addr(IOMMUTLBEntry *iotlb, void **vaddr, return true; } +/* Propagate a guest IOTLB invalidation to the host (nested mode) */ +static void vfio_iommu_unmap_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb) +{ + VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n); + struct vfio_iommu_type1_cache_invalidate ustruct = {}; + VFIOContainer *container = giommu->container; + int ret; + + assert(iotlb->perm == IOMMU_NONE); + + ustruct.argsz = sizeof(ustruct); + ustruct.flags = 0; + ustruct.info.argsz = sizeof(struct iommu_cache_invalidate_info); + ustruct.info.version = IOMMU_CACHE_INVALIDATE_INFO_VERSION_1; + ustruct.info.cache = IOMMU_CACHE_INV_TYPE_IOTLB; + + switch (iotlb->granularity) { + case IOMMU_INV_GRAN_DOMAIN: + ustruct.info.granularity = IOMMU_INV_GRANU_DOMAIN; + break; + case IOMMU_INV_GRAN_PASID: + { + struct iommu_inv_pasid_info *pasid_info; + int archid = -1; + + pasid_info = &ustruct.info.granu.pasid_info; + ustruct.info.granularity = IOMMU_INV_GRANU_PASID; + if (iotlb->flags & IOMMU_INV_FLAGS_ARCHID) { + pasid_info->flags |= IOMMU_INV_ADDR_FLAGS_ARCHID; + archid = iotlb->arch_id; + } + pasid_info->archid = archid; + trace_vfio_iommu_asid_inv_iotlb(archid); + break; + } + case IOMMU_INV_GRAN_ADDR: + { + hwaddr start = iotlb->iova + giommu->iommu_offset; + struct iommu_inv_addr_info *addr_info; + size_t size = iotlb->addr_mask + 1; + int archid = -1; + + addr_info = &ustruct.info.granu.addr_info; + ustruct.info.granularity = IOMMU_INV_GRANU_ADDR; + if (iotlb->leaf) { + addr_info->flags |= IOMMU_INV_ADDR_FLAGS_LEAF; + } + if (iotlb->flags & IOMMU_INV_FLAGS_ARCHID) { + addr_info->flags |= IOMMU_INV_ADDR_FLAGS_ARCHID; + archid = iotlb->arch_id; + } + addr_info->archid = archid; + addr_info->addr = start; + addr_info->granule_size = size; + addr_info->nb_granules = 1; + trace_vfio_iommu_addr_inv_iotlb(archid, start, size, + 1, iotlb->leaf); + break; + }Should we pass a size to host kernel here, even if vSMMU doesn't support RIL or guest kernel doesn't use RIL? It will cause TLBI issue in this scenario: Guest kernel issues a TLBI cmd without "range" (tg = 0) to invalidate a 2M huge page. Then qemu passed the iova and size (4K) to host kernel. Finally, host kernel issues a TLBI cmd with "range" (4K) which can not invalidate the TLB entry of 2M huge page. (pSMMU supports RIL)In that case the guest will loop over all 4K images belonging to the 2M huge page and invalidate each of them. This should turn into qemu notifications for each 4kB page, no? This is totally inefficient, henceThe guest will not loop over all 4K images belonging to the 2M huge page. The iommu_iotlb_gather->pgsize will be 2M, if a page is 2M huge page. The gather->pgsize will be passed to __arm_smmu_tlb_inv_range as "granule": iommu_iotlb_gather_add_page iommu_iotlb_sync domain->ops->iotlb_sync arm_smmu_iotlb_sync arm_smmu_tlb_inv_range_domain __arm_smmu_tlb_inv_range In the above mentioned scenario, guest kernel will issue a TLBI cmd only with "iova" (tg = 0).OK I get it now. In that case I think I should do a TG=0 invalidation on host even if the host does support RIL. Does that sound correct?Yeah, that's what I thought.The trouble is the uapi struct does not convey such info explicitly. Maybe I should use nb_granules = 0 to detect that case.It is troublesome to correct this issue. Using nb_granules = 0 may be a good choice.I think for this use case RIL should be supported by the guest. Thoughts?Yes. If guest supports RIL, the scenario mentioned above does not exist.After further study I really wonder if it is worth supporting the case where the guest does not use range inval. Passing a nb_granules = 0 (~ size) would be OK to perform the TG=0 range invalidation on the host. however host arm_smmu_tlb_inv_range_domain() then calls arm_smmu_atc_inv_domain(smmu_domain, 0, iova, size) which needs a size. We would need to trap guest CMD_ATC_INV and use different code paths on host smmu driver to cascade the various cache invalidations. At the moment, without maintainer giodance, I am a bit reluctant to add this extra complexity. So I would be inclined to fail in QEMU if we detect TG=0 is being used by the guest. Recent guest kernels would be a prerequisite for this use case. What do you think?
Sorry for late reply. How do I determine whether the guest kernel is a recent guest kernel? If I use a modified kernel or an old kernel, this method will cause problems when the vm runs business. I didn't think of the problem of CMD_ATC_INV. In my opinion, it is the best choice to add processing of CMD_ATC_INV. And some function of smmu driver will issues a separate CMD_ATC_INV, for example arm_smmu_enable_ats(). Is it possible that Qemu needs to support the processing of CMD_ATC_INV in the future? I also thought of another not-so-good idea. We can use the granule_size to give CMD_ATC_INV a corresponding maximum page size. This will result in poor performance. For example: granule_size maximum_page_size 4K 1G Thanks, Kunkun Jiang
Thanks EricThanks, Kunkun JiangThanks EricThanks, Kunkun Jiangthe support of RIL on guest side and QEMU device. What do I miss? Thanks EricThanks, Kunkun Jiang+ } + + ret = ioctl(container->fd, VFIO_IOMMU_CACHE_INVALIDATE, &ustruct); + if (ret) { + error_report("%p: failed to invalidate CACHE (%d)", container, ret); + } +} + static void vfio_iommu_map_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb) { VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n); @@ -776,6 +843,35 @@ static void vfio_dma_unmap_ram_section(VFIOContainer *container, } } +static void vfio_prereg_listener_region_add(MemoryListener *listener, + MemoryRegionSection *section) +{ + VFIOContainer *container = + container_of(listener, VFIOContainer, prereg_listener); + Error *err = NULL; + + if (!memory_region_is_ram(section->mr)) { + return; + } + + vfio_dma_map_ram_section(container, section, &err); + if (err) { + error_report_err(err); + } +} +static void vfio_prereg_listener_region_del(MemoryListener *listener, + MemoryRegionSection *section) +{ + VFIOContainer *container = + container_of(listener, VFIOContainer, prereg_listener); + + if (!memory_region_is_ram(section->mr)) { + return; + } + + vfio_dma_unmap_ram_section(container, section); +} + static void vfio_listener_region_add(MemoryListener *listener, MemoryRegionSection *section) { @@ -879,9 +975,10 @@ static void vfio_listener_region_add(MemoryListener *listener, memory_region_ref(section->mr); if (memory_region_is_iommu(section->mr)) { + IOMMUNotify notify; VFIOGuestIOMMU *giommu; IOMMUMemoryRegion *iommu_mr = IOMMU_MEMORY_REGION(section->mr); - int iommu_idx; + int iommu_idx, flags; trace_vfio_listener_region_add_iommu(iova, end); /* @@ -900,8 +997,18 @@ static void vfio_listener_region_add(MemoryListener *listener, llend = int128_sub(llend, int128_one()); iommu_idx = memory_region_iommu_attrs_to_index(iommu_mr,MEMTXATTRS_UNSPECIFIED);- iommu_notifier_init(&giommu->n, vfio_iommu_map_notify, - IOMMU_NOTIFIER_IOTLB_EVENTS, + + if (container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) { + /* IOTLB unmap notifier to propagate guest IOTLB invalidations */ + flags = IOMMU_NOTIFIER_UNMAP; + notify = vfio_iommu_unmap_notify; + } else { + /* MAP/UNMAP IOTLB notifier */ + flags = IOMMU_NOTIFIER_IOTLB_EVENTS; + notify = vfio_iommu_map_notify; + } + + iommu_notifier_init(&giommu->n, notify, flags, section->offset_within_region, int128_get64(llend), iommu_idx); @@ -921,7 +1028,9 @@ static void vfio_listener_region_add(MemoryListener *listener, goto fail; } QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next); - memory_region_iommu_replay(giommu->iommu, &giommu->n); + if (flags & IOMMU_NOTIFIER_MAP) { + memory_region_iommu_replay(giommu->iommu, &giommu->n); + } return; } @@ -1205,10 +1314,16 @@ static const MemoryListener vfio_memory_listener = { .log_sync = vfio_listener_log_sync, }; +static MemoryListener vfio_memory_prereg_listener = { + .region_add = vfio_prereg_listener_region_add, + .region_del = vfio_prereg_listener_region_del, +}; + static void vfio_listener_release(VFIOContainer *container) { memory_listener_unregister(&container->listener); - if (container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU) { + if (container->iommu_type == VFIO_SPAPR_TCE_v2_IOMMU || + container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) { memory_listener_unregister(&container->prereg_listener); } } @@ -1858,6 +1973,20 @@ static int vfio_connect_container(VFIOGroup *group, AddressSpace *as, vfio_get_iommu_info_migration(container, info); } g_free(info); + + if (container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) { + container->prereg_listener = vfio_memory_prereg_listener; + memory_listener_register(&container->prereg_listener, + &address_space_memory); + if (container->error) { + memory_listener_unregister(&container->prereg_listener); + ret = -1; + error_propagate_prepend(errp, container->error, + "RAM memory listener initialization failed " + "for container"); + goto free_container_exit; + } + } break; } case VFIO_SPAPR_TCE_v2_IOMMU: diff --git a/hw/vfio/pci.c b/hw/vfio/pci.c index 5c65aa0a98..cad7deec71 100644 --- a/hw/vfio/pci.c +++ b/hw/vfio/pci.c @@ -2773,6 +2773,25 @@ static void vfio_unregister_req_notifier(VFIOPCIDevice *vdev) vdev->req_enabled = false; } +static int vfio_iommu_set_pasid_table(PCIBus *bus, int32_t devfn, + IOMMUConfig *config) +{ + PCIDevice *pdev = bus->devices[devfn]; + VFIOPCIDevice *vdev = DO_UPCAST(VFIOPCIDevice, pdev, pdev); + VFIOContainer *container = vdev->vbasedev.group->container; + struct vfio_iommu_type1_set_pasid_table info; + + info.argsz = sizeof(info); + info.flags = VFIO_PASID_TABLE_FLAG_SET; + memcpy(&info.config, &config->pasid_cfg, sizeof(config->pasid_cfg)); + + return ioctl(container->fd, VFIO_IOMMU_SET_PASID_TABLE, &info); +} + +static PCIPASIDOps vfio_pci_pasid_ops = { + .set_pasid_table = vfio_iommu_set_pasid_table, +}; + static void vfio_realize(PCIDevice *pdev, Error **errp) { VFIOPCIDevice *vdev = VFIO_PCI(pdev); @@ -3084,6 +3103,8 @@ static void vfio_realize(PCIDevice *pdev, Error **errp) vfio_register_req_notifier(vdev); vfio_setup_resetfn_quirk(vdev); + pci_setup_pasid_ops(pdev, &vfio_pci_pasid_ops); + return; out_deregister: diff --git a/hw/vfio/trace-events b/hw/vfio/trace-events index 936d29d150..43696afc15 100644 --- a/hw/vfio/trace-events +++ b/hw/vfio/trace-events @@ -120,6 +120,8 @@ vfio_region_sparse_mmap_header(const char *name, int index, int nr_areas) "Devic vfio_region_sparse_mmap_entry(int i, unsigned long start, unsigned long end) "sparse entry %d [0x%lx - 0x%lx]" vfio_get_dev_region(const char *name, int index, uint32_t type, uint32_t subtype) "%s index %d, %08x/%0x8" vfio_dma_unmap_overflow_workaround(void) "" +vfio_iommu_addr_inv_iotlb(int asid, uint64_t addr, uint64_t size, uint64_t nb_granules, bool leaf) "nested IOTLB invalidate asid=%d, addr=0x%"PRIx64" granule_size=0x%"PRIx64" nb_granules=0x%"PRIx64" leaf=%d" +vfio_iommu_asid_inv_iotlb(int asid) "nested IOTLB invalidate asid=%d" # platform.c vfio_platform_base_device_init(char *name, int groupid) "%s belongs to group #%d"...