KVM: Never clear KVM_REQ_VM_DEAD from a vCPU's requests
Sean Christopherson
maintainerPATCH 1/1 KVM: Never clear KVM_REQ_VM_DEAD from a vCPU's requests
Use kvm_test_request() instead of kvm_check_request() when querying KVM_REQ_VM_DEAD, i.e. don't clear KVM_REQ_VM_DEAD, as the entire purpose of KVM_REQ_VM_DEAD is to prevent the vCPU from enterring the guest ever again, even if userspace insists on redoing KVM_RUN.
Ensuring KVM_REQ_VM_DEAD is never cleared will allow relaxing KVM's rule that ioctls can't be invoked on dead VMs, to only disallow ioctls if the VM is bugged, i.e. if KVM hit a KVM_BUG_ON().
Opportunistically add compile-time assertions to guard against clearing KVM_REQ_VM_DEAD through the standard APIs.
---
This was previously posted as part of a series to add an ioctl to allow userspace to terminate a TDX VM. I'm posting it separately because it fixes a very real flaw, although the flaw is probably benign in practice? I don't want to find out though, i.e. I don't want to wait to get this merged.
This came back on my radar when fixing a bug in x86's pre-fault code, and looking at that code made it pretty clear that clearing KVM_REQ_VM_DEAD could make a bad situation worse.
arch/arm64/kvm/arm.c | 2 +- arch/x86/kvm/mmu/mmu.c | 2 +- arch/x86/kvm/vmx/tdx.c | 2 +- arch/x86/kvm/x86.c | 2 +- include/linux/kvm_host.h | 9 +++++++-- 5 files changed, 11 insertions(+), 6 deletions(-)
arch/arm64/kvm/arm.c +1 -1
@@ -1108,7 +1108,7 @@ static int kvm_vcpu_suspend(struct kvm_vcpu *vcpu)static int check_vcpu_requests (struct kvm_vcpu *vcpu) {if (kvm_request_pending (vcpu)) { -if (kvm_check_request (KVM_REQ_VM_DEAD , vcpu)) +if (kvm_test_request (KVM_REQ_VM_DEAD , vcpu))return -EIO ;if (kvm_check_request (KVM_REQ_SLEEP , vcpu))
arch/x86/kvm/mmu/mmu.c +1 -1
@@ -5033,7 +5033,7 @@ static int kvm_tdp_page_prefault(struct kvm_vcpu *vcpu, gpa_t gpa,if (signal_pending (current))return -EINTR ; -if (kvm_check_request (KVM_REQ_VM_DEAD , vcpu)) +if (kvm_test_request (KVM_REQ_VM_DEAD , vcpu))return -EIO ;cond_resched ();
arch/x86/kvm/vmx/tdx.c +1 -1
@@ -1995,7 +1995,7 @@ static int tdx_handle_ept_violation(struct kvm_vcpu *vcpu)if (kvm_vcpu_has_events (vcpu) ||signal_pending (current))break ; -if (kvm_check_request (KVM_REQ_VM_DEAD , vcpu)) { +if (kvm_test_request (KVM_REQ_VM_DEAD , vcpu)) { ret = -EIO ;break ; }
arch/x86/kvm/x86.c +1 -1
@@ -8060,7 +8060,7 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)bool req_immediate_exit = false;if (kvm_request_pending (vcpu)) { -if (kvm_check_request (KVM_REQ_VM_DEAD , vcpu)) { +if (kvm_test_request (KVM_REQ_VM_DEAD , vcpu)) { r = -EIO ; goto out; }
include/linux/kvm_host.h +7 -2
@@ -2324,13 +2324,18 @@ static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)return test_bit (req &KVM_REQUEST_MASK , (void *)&vcpu->requests); } -static inline void kvm_clear_request (int req,struct kvm_vcpu *vcpu) +static __always_inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu) { +BUILD_BUG_ON (req ==KVM_REQ_VM_DEAD ); +clear_bit (req &KVM_REQUEST_MASK , (void *)&vcpu->requests); } -static inline bool kvm_check_request (int req,struct kvm_vcpu *vcpu) +static __always_inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu) { +/* Once a VM is dead, it needs to stay dead. */ +BUILD_BUG_ON (req ==KVM_REQ_VM_DEAD ); +if (kvm_test_request (req, vcpu)) {kvm_clear_request (req, vcpu);
base-commit: a806d364ef288a6443a1337820ea8410a7ccc6b3
Signature · 1 line
2.55.0.679.g6767b8d81c-goog
[email protected]
AI reviewThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Pre-existing issues:
- [High] Unhandled KVM_REQ_VM_DEAD across multiple architectures (s390, RISC-V, PowerPC, LoongArch) causes dead VMs to either improperly resume guest execution or get stuck in an infinite kernel loop. --
--- Patch [1]: [PATCH] KVM: Never clear KVM_REQ_VM_DEAD from a vCPU's requests --- commit c8b0ec7047292dc9b84167da2e2a4d9abf600813
Author: Sean Christopherson <[email protected]>
KVM: Never clear KVM_REQ_VM_DEAD from a vCPU's requests
This commit modifies KVM_REQ_VM_DEAD checks to use kvm_test_request() instead of kvm_check_request(), ensuring the flag is never cleared. It also adds a build-time assertion to guard against clearing it via standard APIs.
Quoted 12 lines
include/linux/kvm_host.h +2
-1@@ -2324,13 +2324,18 @@ static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)return test_bit (req &KVM_REQUEST_MASK , (void *)&vcpu->requests); } -static inline void kvm_clear_request (int req,struct kvm_vcpu *vcpu) +static __always_inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu) { +BUILD_BUG_ON (req ==KVM_REQ_VM_DEAD );
include/linux/kvm_host.h static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu) · line 2338
Quoted 10 lines
include/linux/kvm_host.h +5
-1+clear_bit (req &KVM_REQUEST_MASK , (void *)&vcpu->requests); } -static inline bool kvm_check_request (int req,struct kvm_vcpu *vcpu) +static __always_inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu) { +/* Once a VM is dead, it needs to stay dead. */ +BUILD_BUG_ON (req ==KVM_REQ_VM_DEAD ); +
[Severity: High]
This is a pre-existing issue, but while tightening the KVM_REQ_VM_DEAD API, should this request be explicitly handled on other architectures?
Currently, s390, RISC-V, PowerPC, and LoongArch do not check for KVM_REQ_VM_DEAD in their request handling loops. Because the flag is never handled, kvm_request_pending() remains permanently true when a VM is marked dead (e.g., via KVM_BUG_ON() in common pfncache or MMIO paths).
Could this cause the vCPU run loops to infinitely spin inside the kernel without returning to userspace? For example:
- arch/riscv/kvm/vcpu.c:kvm_riscv_check_vcpu_requests()
- arch/powerpc/kvm/powerpc.c:kvm_arch_vcpu_ioctl_run() calling kvmppc_core_check_requests()
- arch/loongarch/kvm/vcpu.c:kvm_check_requests() inside kvm_pre_enter_guest()
Additionally, for s390:
- arch/s390/kvm/kvm-s390.c:kvm_s390_handle_requests() returns 0 when no other requests are handled. Can this cause the dead VM to improperly resume guest execution via sie64a() with a broken host state?
Quoted 3 lines
include/linux/kvm_host.h +0
-0if (kvm_test_request (req, vcpu)) {kvm_clear_request (req, vcpu);
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Quoted 6 lines · Sean Christopherson
Use
kvm_test_request()instead ofkvm_check_request()when queryingKVM_REQ_VM_DEAD, i.e. don't clearKVM_REQ_VM_DEAD, as the entire purpose ofKVM_REQ_VM_DEADis to prevent the vCPU from enterring the guest ever again, even if userspace insists on redoingKVM_RUN.
Quoted 8 lines
Ensuring
KVM_REQ_VM_DEADis never cleared will allow relaxing KVM's rule that ioctls can't be invoked on dead VMs, to only disallow ioctls if the VM is bugged, i.e. if KVM hit aKVM_BUG_ON().Opportunistically add compile-time assertions to guard against clearing
KVM_REQ_VM_DEADthrough the standard APIs.Signed-off-by Sean Christopherson <[email protected]>
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Quoted 5 lines · Sean Christopherson
Use
kvm_test_request()instead ofkvm_check_request()when queryingKVM_REQ_VM_DEAD, i.e. don't clearKVM_REQ_VM_DEAD, as the entire purpose ofKVM_REQ_VM_DEADis to prevent the vCPU from enterring the guest ever again, even if userspace insists on redoingKVM_RUN.
Quoted 8 lines
Ensuring
KVM_REQ_VM_DEADis never cleared will allow relaxing KVM's rule that ioctls can't be invoked on dead VMs, to only disallow ioctls if the VM is bugged, i.e. if KVM hit aKVM_BUG_ON().Opportunistically add compile-time assertions to guard against clearing
KVM_REQ_VM_DEADthrough the standard APIs.Signed-off-by Sean Christopherson <[email protected]>
FWIW, I build tested that both gcc and clang could build successfully (both -O2 and -Os).
[...]
Quoted 13 lines
include/linux/kvm_host.h +3
-1@@ -2324,13 +2324,18 @@ static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu)return test_bit (req &KVM_REQUEST_MASK , (void *)&vcpu->requests); } -static inline void kvm_clear_request (int req,struct kvm_vcpu *vcpu) +static __always_inline void kvm_clear_request(int req, struct kvm_vcpu *vcpu) { +BUILD_BUG_ON (req ==KVM_REQ_VM_DEAD ); +clear_bit (req &KVM_REQUEST_MASK , (void *)&vcpu->requests); }
include/linux/kvm_host.h static inline bool kvm_test_request(int req, struct kvm_vcpu *vcpu) · line 2341
Quoted 10 lines
include/linux/kvm_host.h +4
-1-static inline bool kvm_check_request (int req,struct kvm_vcpu *vcpu) +static __always_inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu) { +/* Once a VM is dead, it needs to stay dead. */ +BUILD_BUG_ON (req ==KVM_REQ_VM_DEAD ); +if (kvm_test_request (req, vcpu)) {kvm_clear_request (req, vcpu);
Nit: AFAICT the change to __always_inline is to avoid build failure. Perhaps explicitly mention this in changelog?
Btw, I also tried building the kernel after reverting __always_inline to plain inline, but indeed got build error when using clang (clang failed with both -O2 and -Os, but gcc was fine for both, though).
Quoted 10 lines · Sean Christopherson
Use
kvm_test_request()instead ofkvm_check_request()when queryingKVM_REQ_VM_DEAD, i.e. don't clearKVM_REQ_VM_DEAD, as the entire purpose ofKVM_REQ_VM_DEADis to prevent the vCPU from enterring the guest ever again, even if userspace insists on redoingKVM_RUN.Ensuring
KVM_REQ_VM_DEADis never cleared will allow relaxing KVM's rule that ioctls can't be invoked on dead VMs, to only disallow ioctls if the VM is bugged, i.e. if KVM hit aKVM_BUG_ON().[...]
Applied to kvm-x86 fixes, thanks!
[1/1] KVM: Never clear KVM_REQ_VM_DEAD from a vCPU's requests
https://github.com/kvm-x86/linux/commit/8cd280282d12