2024-04-12 09:24:07

by Wujie Duan

Subject: [PATCH] KVM: loongarch: Add vcpu id check before create vcpu

Add a pre-allocation arch condition to checks that vcpu id should
smaller than max_vcpus

Signed-off-by: Wujie Duan <[email protected]>
---
arch/loongarch/kvm/vcpu.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/arch/loongarch/kvm/vcpu.c b/arch/loongarch/kvm/vcpu.c
index 3a8779065f73..d41cacf39583 100644
--- a/arch/loongarch/kvm/vcpu.c
+++ b/arch/loongarch/kvm/vcpu.c
@@ -884,6 +884,9 @@ long kvm_arch_vcpu_async_ioctl(struct file *filp,

int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id)
{
+ if (id >= kvm->max_vcpus)
+ return -EINVAL;
+
return 0;
}

--
2.40.1

2024-04-12 18:42:23

by Sean Christopherson

Subject: Re: [PATCH] KVM: loongarch: Add vcpu id check before create vcpu

On Fri, Apr 12, 2024, Wujie Duan wrote:
> Add a pre-allocation arch condition to checks that vcpu id should
> smaller than max_vcpus
>
> Signed-off-by: Wujie Duan <[email protected]>
> ---
> arch/loongarch/kvm/vcpu.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/arch/loongarch/kvm/vcpu.c b/arch/loongarch/kvm/vcpu.c
> index 3a8779065f73..d41cacf39583 100644
> --- a/arch/loongarch/kvm/vcpu.c
> +++ b/arch/loongarch/kvm/vcpu.c
> @@ -884,6 +884,9 @@ long kvm_arch_vcpu_async_ioctl(struct file *filp,
>
> int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id)
> {
> + if (id >= kvm->max_vcpus)
> + return -EINVAL;

Do you have a testcase for this? If I'm following the LoongArch code correctly,
I don't think this is actually necessary.

In arch/loongarch/include/asm/kvm_host.h:

#define KVM_MAX_VCPUS 256

without a #define KVM_MAX_VCPU_IDS in loongarch/, AFAICT. And so the common
code in include/linux/kvm_host.h will do:

#ifndef KVM_MAX_VCPU_IDS
#define KVM_MAX_VCPU_IDS KVM_MAX_VCPUS
#endif

LoongArch's kvm_vm_ioctl_check_extension() reports that to userspace:

case KVM_CAP_MAX_VCPU_ID:
r = KVM_MAX_VCPU_IDS;

and the common kvm_vm_ioctl_create_vcpu() does:

if (id >= KVM_MAX_VCPU_IDS)
return -EINVAL;

and the common kvm_create_vm() does:

kvm->max_vcpus = KVM_MAX_VCPUS;

with again no override of max_vcpus in LoongArch or common KVM. So unless I'm
missing something, manually checking max_vcpus in LoongArch's kvm_arch_vcpu_precreate()
is unnecessary.

2024-04-15 02:10:23

by maobibo

Subject: Re: [PATCH] KVM: loongarch: Add vcpu id check before create vcpu

On 2024/4/13 上午2:41, Sean Christopherson wrote:
> On Fri, Apr 12, 2024, Wujie Duan wrote:
>> Add a pre-allocation arch condition to checks that vcpu id should
>> smaller than max_vcpus
>>
>> Signed-off-by: Wujie Duan <[email protected]>
>> ---
>> arch/loongarch/kvm/vcpu.c | 3 +++
>> 1 file changed, 3 insertions(+)
>>
>> diff --git a/arch/loongarch/kvm/vcpu.c b/arch/loongarch/kvm/vcpu.c
>> index 3a8779065f73..d41cacf39583 100644
>> --- a/arch/loongarch/kvm/vcpu.c
>> +++ b/arch/loongarch/kvm/vcpu.c
>> @@ -884,6 +884,9 @@ long kvm_arch_vcpu_async_ioctl(struct file *filp,
>>
>> int kvm_arch_vcpu_precreate(struct kvm *kvm, unsigned int id)
>> {
>> + if (id >= kvm->max_vcpus)
>> + return -EINVAL;
>
> Do you have a testcase for this? If I'm following the LoongArch code correctly,
> I don't think this is actually necessary.
>
> In arch/loongarch/include/asm/kvm_host.h:
>
> #define KVM_MAX_VCPUS 256
>
> without a #define KVM_MAX_VCPU_IDS in loongarch/, AFAICT. And so the common
> code in include/linux/kvm_host.h will do:
>
> #ifndef KVM_MAX_VCPU_IDS
> #define KVM_MAX_VCPU_IDS KVM_MAX_VCPUS
> #endif
>
> LoongArch's kvm_vm_ioctl_check_extension() reports that to userspace:
>
> case KVM_CAP_MAX_VCPU_ID:
> r = KVM_MAX_VCPU_IDS;
>
> and the common kvm_vm_ioctl_create_vcpu() does:
>
> if (id >= KVM_MAX_VCPU_IDS)
> return -EINVAL;
>
> and the common kvm_create_vm() does:
>
> kvm->max_vcpus = KVM_MAX_VCPUS;
>
> with again no override of max_vcpus in LoongArch or common KVM. So unless I'm
> missing something, manually checking max_vcpus in LoongArch's kvm_arch_vcpu_precreate()
> is unnecessary.
>
yes, you are right. There is already such checking in function
kvm_vm_ioctl_create_vcpu(), and it is unnecessary to add the same
checking in function kvm_arch_vcpu_precreate().

And thanks for pointing it out.

Regards
Bibo Mao