linux-kernel - Re: [PATCH 07/11] KVM: SVM: do not allocate struct svm_cpu

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 9 Nov 2022 15:58:10 +0000
From:   Sean Christopherson <seanjc@...gle.com>
To:     Paolo Bonzini <pbonzini@...hat.com>
Cc:     linux-kernel@...r.kernel.org, kvm@...r.kernel.org,
        thomas.lendacky@....com, jmattson@...gle.com
Subject: Re: [PATCH 07/11] KVM: SVM: do not allocate struct svm_cpu_data
 dynamically

On Wed, Nov 09, 2022, Paolo Bonzini wrote:
> The svm_data percpu variable is a pointer, but it is allocated when
> KVM is loaded (via svm_hardware_setup), not at hardware_enable time.

Parantheses.

> Just allocate room for it statically, that is more efficient and
> does not waste any memory compared to the status quo.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@...hat.com>

Reviewed-by: Sean Christopherson <seanjc@...gle.com>

> @@ -3442,7 +3431,7 @@ static int svm_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
>  
>  static void reload_tss(struct kvm_vcpu *vcpu)
>  {
> -	struct svm_cpu_data *sd = per_cpu(svm_data, vcpu->cpu);
> +	struct svm_cpu_data *sd = per_cpu_ptr(&svm_data, vcpu->cpu);
>  
>  	sd->tss_desc->type = 9; /* available 32/64-bit TSS */
>  	load_TR_desc();
> @@ -3450,7 +3439,7 @@ static void reload_tss(struct kvm_vcpu *vcpu)
>  
>  static void pre_svm_run(struct kvm_vcpu *vcpu)
>  {
> -	struct svm_cpu_data *sd = per_cpu(svm_data, vcpu->cpu);
> +	struct svm_cpu_data *sd = per_cpu_ptr(&svm_data, vcpu->cpu);
>  	struct vcpu_svm *svm = to_svm(vcpu);
>  
>  	/*
> @@ -3919,7 +3908,7 @@ static noinstr void svm_vcpu_enter_exit(struct kvm_vcpu *vcpu)
>  	if (sev_es_guest(vcpu->kvm)) {
>  		__svm_sev_es_vcpu_run(svm);
>  	} else {
> -		struct svm_cpu_data *sd = per_cpu(svm_data, vcpu->cpu);
> +		struct svm_cpu_data *sd = per_cpu_ptr(&svm_data, vcpu->cpu);

At some point we should replace the vcpu->cpu usage with this_cpu_ptr().  All of
the code that does per_cpu_ptr(&svm_data, vcpu->cpu) is doomed if vcpu->cpu isn't
the current CPU.