KVM: x86: VM alloc bug fix and cleanup
cover KVM: x86: VM alloc bug fix and cleanup
Fix a (fairly) long standing NULL pointer dereference if VM allocation fails, and do a bit of clean up on top.
I would have preferred to omit patch 01, i.e. fix the bug via patch 02, but unfortunately (long term support) kernel 4.19 doesn't have the accounting changes, which would make backporting the fix extra annoying for no real benefit.
Sean Christopherson (3):
KVM: x86: Gracefully handle __vmalloc() failure during VM allocation
KVM: x86: Directly return __vmalloc() result in ->vm_alloc()
KVM: x86: Consolidate VM allocation and free for VMX and SVM
arch/x86/include/asm/kvm_host.h | 12 ++++-------- arch/x86/kvm/svm.c | 15 ++++----------- arch/x86/kvm/vmx/vmx.c | 16 ++++------------ arch/x86/kvm/x86.c | 7 +++++++ 4 files changed, 19 insertions(+), 31 deletions(-)
Signature · 1 line
2.24.1
PATCH 1/3 KVM: x86: Gracefully handle __vmalloc() failure during VM allocation
Check the result of __vmalloc() to avoid dereferencing a NULL pointer in the event that allocation failres.
--- arch/x86/kvm/svm.c | 4 ++++ arch/x86/kvm/vmx/vmx.c | 4 ++++ 2 files changed, 8 insertions(+)
arch/x86/kvm/svm.c +4 -0
@@ -1951,6 +1951,10 @@ static struct kvm *svm_vm_alloc(void)struct kvm_svm *kvm_svm =__vmalloc (sizeof (struct kvm_svm ),GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); + +if (!kvm_svm) +return NULL ; +return &kvm_svm->kvm; }
arch/x86/kvm/vmx/vmx.c +4 -0
@@ -6662,6 +6662,10 @@ static struct kvm *vmx_vm_alloc(void)struct kvm_vmx *kvm_vmx =__vmalloc (sizeof (struct kvm_vmx ),GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); + +if (!kvm_vmx) +return NULL ; +return &kvm_vmx->kvm; }
Signature · 1 line
2.24.1
PATCH 2/3 KVM: x86: Directly return __vmalloc() result in ->vm_alloc()
In reply to Sean ChristophersonDirectly return the __vmalloc() result in {svm,vmx}_vm_alloc() to pave the way for handling VM alloc/free in common x86 code, and to obviate the need to check the result of __vmalloc() in vendor specific code. Add a build-time assertion to ensure each structs' "kvm" field stays at offset 0, which allows interpreting a "struct kvm_{svm,vmx}" as a "struct kvm".
--- arch/x86/kvm/svm.c | 12 ++++-------- arch/x86/kvm/vmx/vmx.c | 12 ++++-------- 2 files changed, 8 insertions(+), 16 deletions(-)
arch/x86/kvm/svm.c +4 -8
@@ -1948,19 +1948,15 @@ static void __unregister_enc_region_locked(struct kvm *kvm,static struct kvm *svm_vm_alloc (void ) { -struct kvm_svm *kvm_svm =__vmalloc (sizeof (struct kvm_svm ), -GFP_KERNEL_ACCOUNT | __GFP_ZERO, -PAGE_KERNEL ); +BUILD_BUG_ON (offsetof(struct kvm_svm , kvm) !=0 ); -if (!kvm_svm) -return NULL ; - -return &kvm_svm->kvm; +return __vmalloc (sizeof (struct kvm_svm ), +GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); }static void svm_vm_free (struct kvm *kvm) { -vfree (to_kvm_svm (kvm)); +vfree (kvm); }static void sev_vm_destroy (struct kvm *kvm)
arch/x86/kvm/vmx/vmx.c +4 -8
@@ -6659,20 +6659,16 @@ static void vmx_vcpu_run(struct kvm_vcpu *vcpu)static struct kvm *vmx_vm_alloc (void ) { -struct kvm_vmx *kvm_vmx =__vmalloc (sizeof (struct kvm_vmx ), -GFP_KERNEL_ACCOUNT | __GFP_ZERO, -PAGE_KERNEL ); +BUILD_BUG_ON (offsetof(struct kvm_vmx , kvm) !=0 ); -if (!kvm_vmx) -return NULL ; - -return &kvm_vmx->kvm; +return __vmalloc (sizeof (struct kvm_vmx ), +GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); }static void vmx_vm_free (struct kvm *kvm) {kfree (kvm->arch.hyperv.hv_pa_pg); -vfree (to_kvm_vmx (kvm)); +vfree (kvm); }static void vmx_free_vcpu (struct kvm_vcpu *vcpu)
Signature · 1 line
2.24.1
PATCH 3/3 KVM: x86: Consolidate VM allocation and free for VMX and SVM
In reply to Sean ChristophersonMove the VM allocation and free code to common x86 as the logic is more or less identical across SVM and VMX.
Note, although hyperv.hv_pa_pg is part of the common kvm->arch, it's (currently) only allocated by VMX VMs. But, since kfree() plays nice when passed a NULL pointer, the superfluous call for SVM is harmless and avoids future churn if SVM gains support for HyperV's direct TLB flush.
--- arch/x86/include/asm/kvm_host.h | 12 ++++-------- arch/x86/kvm/svm.c | 13 +++---------- arch/x86/kvm/vmx/vmx.c | 14 +++----------- arch/x86/kvm/x86.c | 7 +++++++ 4 files changed, 17 insertions(+), 29 deletions(-)
arch/x86/include/asm/kvm_host.h +4 -8
@@ -1053,8 +1053,7 @@ struct kvm_x86_ops {bool (*has_emulated_msr)(int index);void (*cpuid_update)(struct kvm_vcpu *vcpu); -struct kvm *(*vm_alloc)(void ); -void (*vm_free)(struct kvm *); +unsigned int (*vm_size)(void );int (*vm_init)(struct kvm *kvm);void (*vm_destroy)(struct kvm *kvm);
@@ -1271,13 +1270,10 @@ extern struct kmem_cache *x86_fpu_cache;#define __KVM_HAVE_ARCH_VM_ALLOCstatic inline struct kvm *kvm_arch_alloc_vm (void ) { -return kvm_x86_ops->vm_alloc (); -} - -static inline void kvm_arch_free_vm (struct kvm *kvm) -{ -return kvm_x86_ops->vm_free (kvm); +return __vmalloc (kvm_x86_ops->vm_size (), +GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); } +void kvm_arch_free_vm (struct kvm *kvm);#define __KVM_HAVE_ARCH_FLUSH_REMOTE_TLBstatic inline int kvm_arch_flush_remote_tlb (struct kvm *kvm)
arch/x86/kvm/svm.c +3 -10
@@ -1946,17 +1946,11 @@ static void __unregister_enc_region_locked(struct kvm *kvm,kfree (region); } -static struct kvm *svm_vm_alloc (void ) +static unsigned int svm_vm_size (void ) {BUILD_BUG_ON (offsetof(struct kvm_svm , kvm) !=0 ); -return __vmalloc (sizeof (struct kvm_svm ), -GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); -} - -static void svm_vm_free (struct kvm *kvm) -{ -vfree (kvm); +return sizeof (struct kvm_svm ); }static void sev_vm_destroy (struct kvm *kvm)
@@ -7385,8 +7379,7 @@ static struct kvm_x86_ops svm_x86_ops __ro_after_init = { .vcpu_free = svm_free_vcpu, .vcpu_reset = svm_vcpu_reset, - .vm_alloc = svm_vm_alloc, - .vm_free = svm_vm_free, + .vm_size = svm_vm_size, .vm_init = svm_vm_init, .vm_destroy = svm_vm_destroy,
arch/x86/kvm/vmx/vmx.c +3 -11
@@ -6657,18 +6657,11 @@ static void vmx_vcpu_run(struct kvm_vcpu *vcpu)vmx_complete_interrupts (vmx); } -static struct kvm *vmx_vm_alloc (void ) +static unsigned int vmx_vm_size (void ) {BUILD_BUG_ON (offsetof(struct kvm_vmx , kvm) !=0 ); -return __vmalloc (sizeof (struct kvm_vmx ), -GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); -} - -static void vmx_vm_free (struct kvm *kvm) -{ -kfree (kvm->arch.hyperv.hv_pa_pg); -vfree (kvm); +return sizeof (struct kvm_vmx ); }static void vmx_free_vcpu (struct kvm_vcpu *vcpu)
@@ -7756,9 +7749,8 @@ static struct kvm_x86_ops vmx_x86_ops __ro_after_init = { .cpu_has_accelerated_tpr = report_flexpriority, .has_emulated_msr = vmx_has_emulated_msr, + .vm_size = vmx_vm_size, .vm_init = vmx_vm_init, - .vm_alloc = vmx_vm_alloc, - .vm_free = vmx_vm_free, .vcpu_create = vmx_create_vcpu, .vcpu_free = vmx_free_vcpu,
arch/x86/kvm/x86.c +7 -0
@@ -9653,6 +9653,13 @@ void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) kvm_x86_ops->sched_in (vcpu, cpu); } +void kvm_arch_free_vm (struct kvm *kvm) +{ +kfree (kvm->arch.hyperv.hv_pa_pg); +vfree (kvm); +} + +int kvm_arch_init_vm (struct kvm *kvm,unsigned long type) {if (type)
Signature · 1 line
2.24.1
Re: [PATCH 3/3] KVM: x86: Consolidate VM allocation and free for VMX and SVM
Quoted 10 lines · Sean Christopherson
Move the VM allocation and free code to common x86 as the logic is more or less identical across SVM and VMX.
Note, although hyperv.hv_pa_pg is part of the common
kvm->arch, it's (currently) only allocated by VMX VMs. But, sincekfree()plays nice when passed a NULL pointer, the superfluous call for SVM is harmless and avoids future churn if SVM gains support for HyperV's direct TLB flush.Signed-off-by Sean Christopherson <[email protected]>
Queued, thanks. Might as well make vm_size a field instead of a pointer to function, sacrificing the BUILD_BUG_ONs:
arch/x86/include/asm/kvm_host.h +2 -2
@@ -1059,7 +1059,7 @@ struct kvm_x86_ops {bool (*has_emulated_msr)(int index);void (*cpuid_update)(struct kvm_vcpu *vcpu); -unsigned int (*vm_size)(void ); +unsigned int vm_size;int (*vm_init)(struct kvm *kvm);void (*vm_destroy)(struct kvm *kvm);
@@ -1276,7 +1276,7 @@ struct kvm_arch_async_pf {#define __KVM_HAVE_ARCH_VM_ALLOCstatic inline struct kvm *kvm_arch_alloc_vm (void ) { -return __vmalloc (kvm_x86_ops->vm_size (), +return __vmalloc (kvm_x86_ops->vm_size,GFP_KERNEL_ACCOUNT | __GFP_ZERO,PAGE_KERNEL ); }void kvm_arch_free_vm (struct kvm *kvm);
arch/x86/kvm/svm.c +1 -8
@@ -1955,13 +1955,6 @@ static void __unregister_enc_region_locked(struct kvm *kvm,kfree (region); } -static unsigned int svm_vm_size (void ) -{ -BUILD_BUG_ON (offsetof(struct kvm_svm , kvm) !=0 ); - -return sizeof (struct kvm_svm ); -} -static void sev_vm_destroy (struct kvm *kvm) {struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info;
@@ -7399,7 +7392,7 @@ static void svm_pre_update_apicv_exec_ctrl(struct kvm *kvm, bool activate) .vcpu_free = svm_free_vcpu, .vcpu_reset = svm_vcpu_reset, - .vm_size = svm_vm_size, + .vm_size = sizeof(struct kvm_svm), .vm_init = svm_vm_init, .vm_destroy = svm_vm_destroy,
arch/x86/kvm/vmx/vmx.c +1 -8
@@ -6645,13 +6645,6 @@ static void vmx_vcpu_run(struct kvm_vcpu *vcpu)vmx_complete_interrupts (vmx); } -static unsigned int vmx_vm_size (void ) -{ -BUILD_BUG_ON (offsetof(struct kvm_vmx , kvm) !=0 ); - -return sizeof (struct kvm_vmx ); -} -static void vmx_free_vcpu (struct kvm_vcpu *vcpu) {struct vcpu_vmx *vmx =to_vmx (vcpu);
@@ -7749,7 +7742,7 @@ static bool vmx_check_apicv_inhibit_reasons(ulong bit) .cpu_has_accelerated_tpr = report_flexpriority, .has_emulated_msr = vmx_has_emulated_msr, - .vm_size = vmx_vm_size, + .vm_size = sizeof(struct kvm_vmx), .vm_init = vmx_vm_init, .vcpu_create = vmx_create_vcpu,
Quoted 8 lines · Sean Christopherson
Fix a (fairly) long standing NULL pointer dereference if VM allocation fails, and do a bit of clean up on top.
I would have preferred to omit patch 01, i.e. fix the bug via patch 02, but unfortunately (long term support) kernel 4.19 doesn't have the accounting changes, which would make backporting the fix extra annoying for no real benefit.
Quoted 5 lines
Sean Christopherson (3):
KVM: x86: Gracefully handle__vmalloc()failure during VM allocation
KVM: x86: Directly return__vmalloc()result in ->vm_alloc()
KVM: x86: Consolidate VM allocation and free for VMX and SVM
Signature · 1 line
Vitaly