Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 1 Apr 2022 02:01:17 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Zeng Guang <guang.zeng@intel.com>
Cc:     Paolo Bonzini <pbonzini@redhat.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>, kvm@vger.kernel.org,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Tony Luck <tony.luck@intel.com>,
        Kan Liang <kan.liang@linux.intel.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        "H. Peter Anvin" <hpa@zytor.com>,
        Kim Phillips <kim.phillips@amd.com>,
        Jarkko Sakkinen <jarkko@kernel.org>,
        Jethro Beekman <jethro@fortanix.com>,
        Kai Huang <kai.huang@intel.com>, x86@kernel.org,
        linux-kernel@vger.kernel.org, Robert Hu <robert.hu@intel.com>,
        Gao Chao <chao.gao@intel.com>
Subject: Re: [PATCH v7 7/8] KVM: x86: Allow userspace set maximum VCPU id for
 VM
Message-ID: <YkZc7cMsDaR5S2hM@google.com>
References: <20220304080725.18135-1-guang.zeng@intel.com>
 <20220304080725.18135-8-guang.zeng@intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220304080725.18135-8-guang.zeng@intel.com>
Precedence: bulk

On Fri, Mar 04, 2022, Zeng Guang wrote:
> Introduce new max_vcpu_id in KVM for x86 architecture. Userspace
> can assign maximum possible vcpu id for current VM session using
> KVM_CAP_MAX_VCPU_ID of KVM_ENABLE_CAP ioctl().
> 
> This is done for x86 only because the sole use case is to guide
> memory allocation for PID-pointer table, a structure needed to
> enable VMX IPI.
> 
> By default, max_vcpu_id set as KVM_MAX_VCPU_IDS.
> 
> Suggested-by: Sean Christopherson <seanjc@google.com>
> Reviewed-by: Maxim Levitsky <mlevitsk@redhat.com>
> Signed-off-by: Zeng Guang <guang.zeng@intel.com>
> ---
>  arch/x86/include/asm/kvm_host.h |  6 ++++++
>  arch/x86/kvm/x86.c              | 11 +++++++++++

The new behavior needs to be documented in api.rst.

>  2 files changed, 17 insertions(+)
> 
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 6dcccb304775..db16aebd946c 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -1233,6 +1233,12 @@ struct kvm_arch {
>  	hpa_t	hv_root_tdp;
>  	spinlock_t hv_root_tdp_lock;
>  #endif
> +	/*
> +	 * VM-scope maximum vCPU ID. Used to determine the size of structures
> +	 * that increase along with the maximum vCPU ID, in which case, using
> +	 * the global KVM_MAX_VCPU_IDS may lead to significant memory waste.
> +	 */
> +	u32 max_vcpu_id;

This should be max_vcpu_ids.  I agree the it _should_ be max_vcpu_id, but KVM's API
for this is awful and we're stuck with the plural name.

>  };
>  
>  struct kvm_vm_stat {
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 4f6fe9974cb5..ca17cc452bd3 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -5994,6 +5994,13 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
>  		kvm->arch.exit_on_emulation_error = cap->args[0];
>  		r = 0;
>  		break;
> +	case KVM_CAP_MAX_VCPU_ID:

I think it makes sense to change kvm_vm_ioctl_check_extension() to return the
current max, it is a VM-scoped ioctl after all.

Amusingly, I think we also need a capability to enumerate that KVM_CAP_MAX_VCPU_ID
is writable.  

> +		if (cap->args[0] <= KVM_MAX_VCPU_IDS) {
> +			kvm->arch.max_vcpu_id = cap->args[0];

This needs to be rejected if kvm->created_vcpus > 0, and that check needs to be
done under kvm_lock, otherwise userspace can bump the max ID after KVM allocates
per-VM structures and trigger buffer overflow.

> +			r = 0;
> +		} else

If-elif-else statements need curly braces for all paths if any path needs braces.
Probably a moot point for this patch due to the above changes.

> +			r = -E2BIG;

This should be -EINVAL, not -E2BIG.

E.g.

	case KVM_CAP_MAX_VCPU_ID:
		r = -EINVAL;
		if (cap->args[0] > KVM_MAX_VCPU_IDS)
			break;

		mutex_lock(&kvm->lock);
		if (!kvm->created_vcpus) {
			kvm->arch.max_vcpu_id = cap->args[0];
			r = 0;
		}
		mutex_unlock(&kvm->lock);
		break;


> +		break;
>  	default:
>  		r = -EINVAL;
>  		break;
> @@ -11067,6 +11074,9 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
>  	struct page *page;
>  	int r;
>  
> +	if (vcpu->vcpu_id >= vcpu->kvm->arch.max_vcpu_id)
> +		return -E2BIG;

Same here, it should be -EINVAL.

> +
>  	vcpu->arch.last_vmentry_cpu = -1;
>  	vcpu->arch.regs_avail = ~0;
>  	vcpu->arch.regs_dirty = ~0;
> @@ -11589,6 +11599,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
>  	spin_lock_init(&kvm->arch.hv_root_tdp_lock);
>  	kvm->arch.hv_root_tdp = INVALID_PAGE;
>  #endif
> +	kvm->arch.max_vcpu_id = KVM_MAX_VCPU_IDS;
>  
>  	INIT_DELAYED_WORK(&kvm->arch.kvmclock_update_work, kvmclock_update_fn);
>  	INIT_DELAYED_WORK(&kvm->arch.kvmclock_sync_work, kvmclock_sync_fn);
> -- 
> 2.27.0
>