2020-04-13 09:18:45

by Paolo Bonzini

Subject: [PATCH] KVM: SVM: fix compilation with modular PSP and non-modular KVM

Use svm_sev_enabled() in order to cull all calls to PSP code. Otherwise,
compilation fails with undefined symbols if the PSP device driver is compiled
as a module and KVM is not.

Reported-by: Uros Bizjak <[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>
---
arch/x86/kvm/svm/sev.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
index 0e3fc311d7da..364ffe32139c 100644
--- a/arch/x86/kvm/svm/sev.c
+++ b/arch/x86/kvm/svm/sev.c
@@ -1117,7 +1117,7 @@ int __init sev_hardware_setup(void)
/* Maximum number of encrypted guests supported simultaneously */
max_sev_asid = cpuid_ecx(0x8000001F);

- if (!max_sev_asid)
+ if (!svm_sev_enabled())
return 1;

/* Minimum ASID value that should be used for SEV guest */
@@ -1156,6 +1156,9 @@ int __init sev_hardware_setup(void)

void sev_hardware_teardown(void)
{
+ if (!svm_sev_enabled())
+ return;
+
bitmap_free(sev_asid_bitmap);
bitmap_free(sev_reclaim_asid_bitmap);

--
2.18.2

2020-04-15 06:09:23

by Sean Christopherson

Subject: Re: [PATCH] KVM: SVM: fix compilation with modular PSP and non-modular KVM

On Mon, Apr 13, 2020 at 03:50:31AM -0400, Paolo Bonzini wrote:
> Use svm_sev_enabled() in order to cull all calls to PSP code. Otherwise,
> compilation fails with undefined symbols if the PSP device driver is compiled
> as a module and KVM is not.
>
> Reported-by: Uros Bizjak <[email protected]>
> Signed-off-by: Paolo Bonzini <[email protected]>
> ---
> arch/x86/kvm/svm/sev.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index 0e3fc311d7da..364ffe32139c 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -1117,7 +1117,7 @@ int __init sev_hardware_setup(void)
> /* Maximum number of encrypted guests supported simultaneously */
> max_sev_asid = cpuid_ecx(0x8000001F);
>
> - if (!max_sev_asid)
> + if (!svm_sev_enabled())
> return 1;
>
> /* Minimum ASID value that should be used for SEV guest */
> @@ -1156,6 +1156,9 @@ int __init sev_hardware_setup(void)
>
> void sev_hardware_teardown(void)
> {
> + if (!svm_sev_enabled())
> + return;
> +

Tabs instead of spaces. Checkpatch also whinges about going past 75 chars
in the changelog.

> bitmap_free(sev_asid_bitmap);
> bitmap_free(sev_reclaim_asid_bitmap);
>
> --
> 2.18.2
>

2020-04-16 20:26:46

by Paolo Bonzini

Subject: Re: [PATCH] KVM: SVM: fix compilation with modular PSP and non-modular KVM

On 16/04/20 15:57, Tom Lendacky wrote:
> On 4/13/20 2:50 AM, Paolo Bonzini wrote:
>> Use svm_sev_enabled() in order to cull all calls to PSP code.  Otherwise,
>> compilation fails with undefined symbols if the PSP device driver is
>> compiled
>> as a module and KVM is not.
>
> The Kconfig support will set CONFIG_KVM_AMD_SEV to "n" in this
> situation, so it might be worth seeing if sev.o could be removed from
> the build at that point. I'll try and look at that when I get a chance,
> but I'm currently buried with a ton of other work.

It could be made to work, but you would have to add stubs to sev.h.

Paolo

2020-04-16 20:32:37

by Tom Lendacky

Subject: Re: [PATCH] KVM: SVM: fix compilation with modular PSP and non-modular KVM

On 4/13/20 2:50 AM, Paolo Bonzini wrote:
> Use svm_sev_enabled() in order to cull all calls to PSP code. Otherwise,
> compilation fails with undefined symbols if the PSP device driver is compiled
> as a module and KVM is not.

The Kconfig support will set CONFIG_KVM_AMD_SEV to "n" in this situation,
so it might be worth seeing if sev.o could be removed from the build at
that point. I'll try and look at that when I get a chance, but I'm
currently buried with a ton of other work.

>
> Reported-by: Uros Bizjak <[email protected]>
> Signed-off-by: Paolo Bonzini <[email protected]>
> ---
> arch/x86/kvm/svm/sev.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
> index 0e3fc311d7da..364ffe32139c 100644
> --- a/arch/x86/kvm/svm/sev.c
> +++ b/arch/x86/kvm/svm/sev.c
> @@ -1117,7 +1117,7 @@ int __init sev_hardware_setup(void)
> /* Maximum number of encrypted guests supported simultaneously */
> max_sev_asid = cpuid_ecx(0x8000001F);
>
> - if (!max_sev_asid)
> + if (!svm_sev_enabled())

It looks like these are spaces instead of tabs, could just be my email
veiwer, though.

> return 1;
>
> /* Minimum ASID value that should be used for SEV guest */
> @@ -1156,6 +1156,9 @@ int __init sev_hardware_setup(void)
>
> void sev_hardware_teardown(void)
> {
> + if (!svm_sev_enabled())
> + return;
> +

Ditto on the spaces/tabs thing, here.

Thanks,
Tom

> bitmap_free(sev_asid_bitmap);
> bitmap_free(sev_reclaim_asid_bitmap);
>
>