[6.6] [Feature] : Intel TDX Backport for OpenAnolis devel-6.6 Kernel — Patchset#2 tdx kvm 6.12 dependence · Pull Request !10547 · anolis/cloud-kernel

bugzilla: https://bugzilla.openanolis.cn/show_bug.cgi?id=40036
Intel TDX Host Module Initialization Backport for OpenAnolis devel-6.6 Kernel

Patch List:
```
anolis: configs: enable CONFIG_INTEL_TDX_HOST on x86
KVM: x86: Return the VM's configured APIC bus frequency when queried
KVM: x86/mmu: Fixup comments missed by the REMOVED_SPTE=>FROZEN_SPTE rename
KVM: selftests: Add test for configure of x86 APIC bus frequency
KVM: selftests: Add guest udelay() utility for x86
KVM: x86: Add a capability to configure bus frequency for APIC timer
KVM: x86: Make nanoseconds per APIC bus cycle a VM variable
KVM: x86: hyper-v: Calculate APIC bus frequency for Hyper-V
KVM: selftests: x86: Add test for KVM_PRE_FAULT_MEMORY
KVM: x86/mmu: Bump pf_taken stat only in the "real" page fault handler
KVM: x86/tdp_mmu: Take a GFN in kvm_tdp_mmu_fast_pf_get_last_sptep()
KVM: x86/tdp_mmu: Rename REMOVED_SPTE to FROZEN_SPTE
KVM: guest_memfd: Use AS_INACCESSIBLE when creating guest_memfd inode
KVM: x86/tdp_mmu: Sprinkle __must_check
KVM: x86: Ensure a full memory barrier is emitted in the VM-Exit path
srcu: Add an API for a memory barrier after SRCU read lock
KVM: VMX: Drop support for forcing UC memory when guest CR0.CD=1
KVM: x86: Remove VMX support for virtualizing guest MTRR memtypes
KVM: x86/mmu: Avoid reacquiring RCU if TDP MMU fails to allocate an SP
KVM: x86/mmu: Unnest TDP MMU helpers that allocate SPs for eager splitting
KVM: x86/mmu: Hard code GFP flags for TDP MMU eager split allocations
KVM: x86/mmu: Always drop mmu_lock to allocate TDP MMU SPs for eager splitting
KVM: x86/mmu: Don't save mmu_invalidate_seq after checking private attr
KVM: x86: Disable KVM_INTEL_PROVE_VE by default
KVM: VMX: Enumerate EPT Violation #VE support in /proc/cpuinfo
KVM: x86/mmu: Print SPTEs on unexpected #VE
KVM: VMX: Dump VMCS on unexpected #VE
KVM: x86/mmu: Add sanity checks that KVM doesn't create EPT #VE SPTEs
KVM: nVMX: Always handle #VEs in L0 (never forward #VEs from L2 to L1)
KVM: nVMX: Initialize #VE info page for vmcs02 when proving #VE support
KVM: VMX: Don't kill the VM on an unexpected #VE
KVM: x86/mmu: Use SHADOW_NONPRESENT_VALUE for atomic zap in TDP MMU
KVM: x86: Move nEPT exit_qualification field from kvm_vcpu_arch to x86_exception
KVM: nVMX: Clear EXIT_QUALIFICATION when injecting an EPT Misconfig
KVM: x86: Explicitly zero kvm_caps during vendor module load
KVM: x86: Fully re-initialize supported_mce_cap on vendor module load
KVM: x86/mmu: Sanity check that __kvm_faultin_pfn() doesn't create noslot pfns
KVM: x86/mmu: Initialize kvm_page_fault's pfn and hva to error values
KVM: x86/mmu: Set kvm_page_fault.hva to KVM_HVA_ERR_BAD for "no slot" faults
KVM: x86/mmu: Handle no-slot faults at the beginning of kvm_faultin_pfn()
KVM: x86/mmu: Move slot checks from __kvm_faultin_pfn() to kvm_faultin_pfn()
KVM: x86/mmu: Explicitly disallow private accesses to emulated MMIO
KVM: x86/mmu: Don't force emulation of L2 accesses to non-APIC internal slots
KVM: x86/mmu: Move private vs. shared check above slot validity checks
KVM: x86/mmu: WARN and skip MMIO cache on private, reserved page faults
KVM: x86/mmu: WARN if upper 32 bits of legacy #PF error code are non-zero
KVM: x86/mmu: Exit to userspace with -EFAULT if private fault hits emulation
KVM: VMX: Introduce test mode related to EPT violation VE
KVM, x86: add architectural support code for #VE
KVM: x86/mmu: Track shadow MMIO value on a per-VM basis
KVM: x86/mmu: Add Suppress VE bit to EPT shadow_mmio_mask/shadow_present_mask
KVM: x86/mmu: Allow non-zero value for non-present SPTE and removed SPTE
KVM: x86/mmu: Replace hardcoded value 0 for the initial value for SPTE
KVM: Allow page-sized MMU caches to be initialized with custom 64-bit values
KVM: VMX: Modify NMI and INTR handlers to take intr_info as function argument
KVM: VMX: Move out vmx_x86_ops to 'main.c' to dispatch VMX and TDX
KVM: x86: Split core of hypercall emulation to helper function
KVM: fix documentation for KVM_CREATE_GUEST_MEMFD
KVM: x86/mmu: Fix a largely theoretical race in kvm_mmu_track_write()
KVM: x86/mmu: Process atomically-zapped SPTEs after TLB flush
KVM: selftests: Drop @selector from segment helpers
KVM: selftests: Init x86's segments during VM creation
KVM: selftests: Add macro for TSS selector, rename up code/data macros
KVM: selftests: Allocate x86's TSS at VM creation
KVM: selftests: Fold x86's descriptor tables helpers into vcpu_init_sregs()
KVM: selftests: Drop superfluous switch() on vm->mode in vcpu_init_sregs()
KVM: selftests: Allocate x86's GDT during VM creation
KVM: selftests: Map x86's exception_handlers at VM creation, not vCPU setup
KVM: selftests: Init IDT and exception handlers for all VMs/vCPUs on x86
KVM: selftests: Rename x86's vcpu_setup() to vcpu_init_sregs()
KVM: selftests: Move x86's descriptor table helpers "up" in processor.c
KVM: selftests: Explicitly clobber the IDT in the "delete memslot" testcase
KVM: selftests: Rework platform_info_test to actually verify #GP
KVM: selftests: Move platform_info_test's main assert into guest code
KVM: selftests: Fix off-by-one initialization of GDT limit
KVM: selftests: Move GDT, IDT, and TSS fields to x86's kvm_vm_arch
KVM: sefltests: Add kvm_util_types.h to hold common types, e.g. vm_vaddr_t
Revert "kvm: selftests: move base kvm_util.h declarations to kvm_util_base.h"
KVM: selftests: Move setting a vCPU's entry point to a dedicated API
KVM: x86/mmu: Precisely invalidate MMU root_role during CPUID update
KVM: x86/mmu: x86: Don't overflow lpage_info when checking attributes
KVM: VMX: Report up-to-date exit qualification to userspace
kvm/x86: allocate the write-tracking metadata on-demand
KVM: x86/mmu: Free TDP MMU roots while holding mmy_lock for read
KVM: x86/mmu: Alloc TDP MMU roots while holding mmu_lock for read
KVM: x86/mmu: Check for usable TDP MMU root while holding mmu_lock for read
KVM: x86/mmu: Skip invalid TDP MMU roots when write-protecting SPTEs
KVM: x86/mmu: Skip invalid roots when zapping leaf SPTEs for GFN range
KVM: x86/mmu: Allow passing '-1' for "all" as_id for TDP MMU iterators
KVM: x86/mmu: Don't do TLB flush when zappings SPTEs in invalid roots
KVM: x86/mmu: Zap invalidated TDP MMU roots at 4KiB granularity
KVM: x86: Drop superfluous check on direct MMU vs. WRITE_PF_TO_SP flag
KVM: x86: Drop dedicated logic for direct MMUs in reexecute_instruction()
KVM: x86/mmu: Don't acquire mmu_lock when using indirect_shadow_pages as a heuristic
KVM: x86/mmu: Retry fault before acquiring mmu_lock if mapping is changing
KVM: selftests: Add a testcase to verify GUEST_MEMFD and READONLY are exclusive
KVM: x86: Update KVM_SW_PROTECTED_VM docs to make it clear they're a WIP
KVM: x86/mmu: always take tdp_mmu_pages_lock
KVM: x86/mmu: remove unnecessary "bool shared" argument from iterators
KVM: x86/mmu: remove unnecessary "bool shared" argument from functions
KVM: x86: Don't sync user-written TSC against startup values
KVM: VMX: drop IPAT in memtype when CD=1 for KVM_X86_QUIRK_CD_NW_CLEARED
KVM: x86/mmu: Zap KVM TDP when noncoherent DMA assignment starts/stops
KVM: x86/mmu: Zap SPTEs on MTRR update iff guest MTRRs are honored
KVM: x86/mmu: Zap SPTEs when CR0.CD is toggled iff guest MTRRs are honored
KVM: x86/mmu: Add helpers to return if KVM honors guest MTRRs

```

--- ---
 **Patch Series Overview:** 
This is a set of 105 patches for the KVM x86 MMU and VMX subsystem backport to the OpenAnolis 6.6 kernel. The changes focus on five major areas: (1) EPT Violation #VE infrastructure for TDX, (2) TDP MMU scalability and correctness improvements, (3) MTRR/memtype handling modernisation, (4) APIC timer bus frequency configurability, and (5) KVM selftest infrastructure enhancements. These patches align the Anolis 6.6 kernel with the upstream prerequisites for TDX host-side enablement while maintaining backward compatibility.

--- ---
 **Key Bug Fixes:**

**Race in kvm_mmu_track_write():**  Fixed a largely theoretical race where a stale SPTE could be observed during concurrent write-tracking, potentially causing incorrect emulation.
lpage_info overflow when checking attributes: Fixed a buffer overflow when validating memory attributes against large page information arrays.
 **Exit qualification reporting to userspace:**  VMX now reports the up-to-date exit_qualification to userspace instead of a potentially stale cached value.
 **mmu_invalidate_seq ordering for private attributes:**  Removed an incorrect save of mmu_invalidate_seq after checking private attributes, avoiding a TOCTOU race in the page fault path.
 **FROZEN_SPTE rename fixup:** Completed the REMOVED_SPTE→FROZEN_SPTE rename in comments that were missed by the initial rename commit.
 **TSC synchronisation on user writes:**  Stopped KVM from incorrectly syncing user-written TSC values against startup values, preventing timestamp jumps.
 **IPAT memtype when CD=1:** Dropped the incorrect IPAT memtype in EPT entries when guest CR0.CD=1 with KVM_X86_QUIRK_CD_NW_CLEARED active.
 **Private fault hitting emulation:**  KVM now exits to userspace with -EFAULT if a private page fault triggers instruction emulation, rather than silently corrupting state.
 **Atomically-zapped SPTEs processed after TLB flush:**  Fixed ordering so that SPTEs zapped atomically are processed only after the required TLB flush completes.
 **guest_memfd documentation:**  Fixed inaccurate documentation for KVM_CREATE_GUEST_MEMFD ioctl parameters.

--- ---
 **Feature Additions:**

**EPT Violation #VE support:**  Added the complete infrastructure for EPT Violation Virtualization Exception (#VE) handling — including Suppress VE bit management in shadow_mmio_mask/shadow_present_mask, #VE info page initialisation for vmcs02, L0 interception of L2 #VEs, VMCS/SPTE dumping on unexpected #VEs, test mode for controlled #VE injection, and /proc/cpuinfo enumeration.
 **Non-zero SHADOW_NONPRESENT_VALUE:**  Enabled configurable initial values for non-present and removed SPTEs, required by TDX where the Suppress VE bit must be set in all non-present entries.
 **APIC timer bus frequency configuration:**  Added per-VM APIC bus cycle variable (KVM_CAP_X86_APIC_BUS_FREQUENCY) to allow userspace to configure the APIC timer bus frequency, with correct Hyper-V frequency calculation.
 **KVM_PRE_FAULT_MEMORY support in selftests:**  Added x86 selftest coverage for the KVM_PRE_FAULT_MEMORY ioctl.
 **guest_memfd AS_INACCESSIBLE:**  Marked guest_memfd inodes with AS_INACCESSIBLE address space flag for proper isolation.
 **VMX/TDX dispatch split:**  Moved vmx_x86_ops to a dedicated main.c file, enabling clean dispatch between VMX and TDX code paths.
 **Hypercall emulation refactoring:**  Split the core of hypercall emulation into a helper function (__kvm_emulate_hypercall) reusable by TDX's TDVMCALL handling.
 **write-tracking metadata on-demand allocation:**  Allocate write-tracking metadata only when needed, reducing memory overhead for VMs that don't use shadow paging.

--- ---
 **Logic & Performance Improvements:**

**TDP MMU root lifecycle under mmu_lock for read:**  Converted TDP MMU root allocation, freeing, and usability checks to operate under mmu_lock for read (not write), significantly reducing lock contention on large VMs.
 **TDP MMU invalid root handling:**  Skip invalid roots during write-protection, GFN-range zapping, and TLB flushes. Zap invalidated roots at 4KiB granularity for faster teardown.
 **TDP MMU iterator simplification:**  Removed unnecessary "bool shared" argument from TDP MMU iterators and helper functions after the mmu_lock refactoring.
 **Eager split allocation cleanup:**  Hard-coded GFP flags, unnested allocation helpers, and always drop mmu_lock before allocating SPs for eager splitting — removing complex conditional paths.
 **MTRR/memtype modernisation:**  Added kvm_honors_guest_mtrrs() helpers, conditional SPTE zapping only when guest MTRRs are actually honored, removed dead VMX memtype virtualization code, and dropped UC memory forcing on CR0.CD=1.
 **Page fault path optimisation:**  Retry faults before acquiring mmu_lock if the mapping is changing; skip mmu_lock acquisition when indirect_shadow_pages is used as a heuristic; handle no-slot faults early to avoid unnecessary work.
 **Private memory fault path hardening:**  Moved private-vs-shared checks above slot validity checks, explicitly disallowed private accesses to emulated MMIO, and added WARN for private faults hitting reserved pages.
 **Memory barrier after SRCU read lock:**  Added srcu_read_lock_memory_barrier() API and used it in the VM-Exit path to ensure a full memory barrier is emitted.

--- --- 
 **Selftest Coverage Added:**

KVM_PRE_FAULT_MEMORY test for x86
APIC bus frequency configuration test
Guest udelay() utility for timing-dependent x86 tests
GUEST_MEMFD and READONLY exclusivity verification
Platform_info_test reworked to verify #GP delivery in guest code
x86 descriptor table (GDT/IDT/TSS) infrastructure refactored for per-VM creation
Exception handler and IDT initialised for all VMs/vCPUs automatically
Separate vCPU entry point API (vcpu_args_set decoupled from init)

--- ---
 **Robustness & Safety:**

All EPT #VE paths include sanity checks preventing accidental #VE injection via invalid SPTEs.
Nested VMX correctly intercepts #VE in L0 and never forwards to L1.
KVM_INTEL_PROVE_VE disabled by default to avoid unexpected VM kills in production.
kvm_page_fault struct fields initialised to error values to catch missing assignments.
__must_check annotations added to TDP MMU helpers that can fail.
KVM capabilities (kvm_caps) fully re-initialised on vendor module reload to prevent stale state

--- ---
 **Testing:** 
The test setup and test cases are documented in the Bugzilla issue .

类型	指派人员	状态
审查	Tianjia Zhang xunlei Dust.Li 半叶枫尘 winterddd xueshuai guobin maqiao-alibaba baolin.wang 冯光辉 David Guixin Liu Xuan Zhuo XiangZao fangyu0809 高家骐	已完成 (0/0人)
测试	小龙已测试通过	已完成 (1/1人)

GVPanolis / cloud-kernel

搜索帮助

GVP anolis / cloud-kernel