Zheyun Shen <szy0127@xxxxxxxxxxx> writes:
> On AMD CPUs without ensuring cache consistency, each memory page
> reclamation in an SEV guest triggers a call to wbinvd_on_all_cpus(),
> thereby affecting the performance of other programs on the host.
>
> Typically, an AMD server may have 128 cores or more, while the SEV guest
> might only utilize 8 of these cores. Meanwhile, host can use qemu-affinity
> to bind these 8 vCPUs to specific physical CPUs.
>
> Therefore, keeping a record of the physical core numbers each time a vCPU
> runs can help avoid flushing the cache for all CPUs every time.
>
> Suggested-by: Sean Christopherson <seanjc@xxxxxxxxxx>
> Signed-off-by: Zheyun Shen <szy0127@xxxxxxxxxxx>
> ---
> arch/x86/kvm/svm/sev.c | 39 ++++++++++++++++++++++++++++++++++++---
> arch/x86/kvm/svm/svm.c | 2 ++
> arch/x86/kvm/svm/svm.h | 5 ++++-
> 3 files changed, 42 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index 1ce67de9d..91469edd1 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -252,6 +252,36 @@ static void sev_asid_free(struct kvm_sev_info *sev)
> sev->misc_cg = NULL;
> }
>
> +static struct cpumask *sev_get_wbinvd_dirty_mask(struct kvm *kvm)
> +{
> + struct kvm_sev_info *sev = &to_kvm_svm(kvm)->sev_info;
There is a helper to get sev_info: to_kvm_sev_info(), if you use that,
sev_get_wbinvd_dirty_mask() helper will not be needed.
> +
> + return sev->wbinvd_dirty_mask;
> +}
> +
> +void sev_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
> +{
> + /*
> + * To optimize cache flushes when memory is reclaimed from an SEV VM,
> + * track physical CPUs that enter the guest for SEV VMs and thus can
> + * have encrypted, dirty data in the cache, and flush caches only for
> + * CPUs that have entered the guest.
> + */
> + cpumask_set_cpu(cpu, sev_get_wbinvd_dirty_mask(vcpu->kvm));
> +}
> +
> +static void sev_do_wbinvd(struct kvm *kvm)
> +{
> + struct cpumask *dirty_mask = sev_get_wbinvd_dirty_mask(kvm);
> +
> + /*
> + * TODO: Clear CPUs from the bitmap prior to flushing. Doing so
> + * requires serializing multiple calls and having CPUs mark themselves
> + * "dirty" if they are currently running a vCPU for the VM.
> + */
> + wbinvd_on_many_cpus(dirty_mask);
> +}
Something like the below
void sev_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
{
/* ... */
cpumask_set_cpu(cpu, to_kvm_sev_info(kvm)->wbinvd_dirty_mask);
}
static void sev_do_wbinvd(struct kvm *kvm)
{
/* ... */
wbinvd_on_many_cpus(to_kvm_sev_info(kvm)->wbinvd_dirty_mask);
}
Regards,
Nikunj
