Modify following APIs for x86 implementation:
1) kvm_arch_main : Query the cpu vendor and cache the value in advance.
2) kvm_arch_post_vm_load: Populate cpu type in the guest memory so that
guest doesn't need to execute cpuid instruction again.
Suggested-by: Sean Christopherson <[email protected]>
Signed-off-by: Vishal Annapurve <[email protected]>
---
.../testing/selftests/kvm/lib/x86_64/processor.c | 15 ++++++++++++---
1 file changed, 12 insertions(+), 3 deletions(-)
diff --git a/tools/testing/selftests/kvm/lib/x86_64/processor.c b/tools/testing/selftests/kvm/lib/x86_64/processor.c
index e22cfc4bf284..ac104653ab43 100644
--- a/tools/testing/selftests/kvm/lib/x86_64/processor.c
+++ b/tools/testing/selftests/kvm/lib/x86_64/processor.c
@@ -19,6 +19,7 @@
#define MAX_NR_CPUID_ENTRIES 100
vm_vaddr_t exception_handlers;
+static int is_cpu_amd = -1;
static void regs_dump(FILE *stream, struct kvm_regs *regs, uint8_t indent)
{
@@ -1019,7 +1020,7 @@ static bool cpu_vendor_string_is(const char *vendor)
bool is_intel_cpu(void)
{
- return cpu_vendor_string_is("GenuineIntel");
+ return (is_cpu_amd == 0);
}
/*
@@ -1027,7 +1028,7 @@ bool is_intel_cpu(void)
*/
bool is_amd_cpu(void)
{
- return cpu_vendor_string_is("AuthenticAMD");
+ return (is_cpu_amd == 1);
}
void kvm_get_cpu_address_width(unsigned int *pa_bits, unsigned int *va_bits)
@@ -1182,9 +1183,15 @@ uint64_t kvm_hypercall(uint64_t nr, uint64_t a0, uint64_t a1, uint64_t a2,
{
uint64_t r;
- asm volatile("vmcall"
+ if (is_amd_cpu())
+ asm volatile("vmmcall"
: "=a"(r)
: "a"(nr), "b"(a0), "c"(a1), "d"(a2), "S"(a3));
+ else
+ asm volatile("vmcall"
+ : "=a"(r)
+ : "a"(nr), "b"(a0), "c"(a1), "d"(a2), "S"(a3));
+
return r;
}
@@ -1314,8 +1321,10 @@ bool vm_is_unrestricted_guest(struct kvm_vm *vm)
void kvm_arch_main(void)
{
+ is_cpu_amd = cpu_vendor_string_is("AuthenticAMD") ? 1 : 0;
}
void kvm_arch_post_vm_load(struct kvm_vm *vm)
{
+ sync_global_to_guest(vm, is_cpu_amd);
}
--
2.37.2.789.g6183377224-goog
On Thu, Sep 08, 2022 at 03:54:41PM -0700, David Matlack wrote:
> Please use "KVM: selftest: ..." for the shortlog.
Doh. "KVM: selftests: ..."
Please use "KVM: selftest: ..." for the shortlog.
On Sat, Sep 03, 2022 at 01:28:47AM +0000, Vishal Annapurve wrote:
> Modify following APIs for x86 implementation:
> 1) kvm_arch_main : Query the cpu vendor and cache the value in advance.
> 2) kvm_arch_post_vm_load: Populate cpu type in the guest memory so that
> guest doesn't need to execute cpuid instruction again.
This commit message only describes a subset of the changes in this
commit, and does not provide any context on why the changes are being
made (other than a clue about avoiding CPUID).
I also think this could be split up into 2 separate commits.
I would suggest first patch changes is_{intel,amd}_cpu() to return a cached
result. e.g.
KVM: selftests: Precompute the result for is_{intel,amd}_cpu()
Cache the vendor CPU type in a global variable so that multiple calls
to is_intel_cpu() do not need to re-execute CPUID. This will be useful
in a follow-up commit to check if running on AMD or Intel from within
a selftest guest where executing CPUID requires a VM-exit.
Then add support for AMD to kvm_hypercall():
KVM: selftests: Add AMD support to kvm_hypercall()
Add support for AMD hypercalls to kvm_hypercall() so that it can be
used in selftests running on Intel or AMD hosts. This will be used in
a follow up commit to ...
As part of this change, sync the global variable is_cpu_amd into the
guest so the guest can determine which hypercall instruction to use
without needing to re-execute CPUID for every hypercall.
>
> Suggested-by: Sean Christopherson <[email protected]>
> Signed-off-by: Vishal Annapurve <[email protected]>
> ---
> .../testing/selftests/kvm/lib/x86_64/processor.c | 15 ++++++++++++---
> 1 file changed, 12 insertions(+), 3 deletions(-)
>
> diff --git a/tools/testing/selftests/kvm/lib/x86_64/processor.c b/tools/testing/selftests/kvm/lib/x86_64/processor.c
> index e22cfc4bf284..ac104653ab43 100644
> --- a/tools/testing/selftests/kvm/lib/x86_64/processor.c
> +++ b/tools/testing/selftests/kvm/lib/x86_64/processor.c
> @@ -19,6 +19,7 @@
> #define MAX_NR_CPUID_ENTRIES 100
>
> vm_vaddr_t exception_handlers;
> +static int is_cpu_amd = -1;
Should this just be a bool? Since you are initializing it before main(),
there is really no way for any code to observe it's pre-initialized
value. And nothing even checks if is_cpu_amd -1, it just silently
returns false from is_intel_cpu() and is_amd_cpu().
>
> static void regs_dump(FILE *stream, struct kvm_regs *regs, uint8_t indent)
> {
> @@ -1019,7 +1020,7 @@ static bool cpu_vendor_string_is(const char *vendor)
>
> bool is_intel_cpu(void)
> {
> - return cpu_vendor_string_is("GenuineIntel");
> + return (is_cpu_amd == 0);
> }
>
> /*
> @@ -1027,7 +1028,7 @@ bool is_intel_cpu(void)
> */
> bool is_amd_cpu(void)
> {
> - return cpu_vendor_string_is("AuthenticAMD");
> + return (is_cpu_amd == 1);
> }
>
> void kvm_get_cpu_address_width(unsigned int *pa_bits, unsigned int *va_bits)
> @@ -1182,9 +1183,15 @@ uint64_t kvm_hypercall(uint64_t nr, uint64_t a0, uint64_t a1, uint64_t a2,
> {
> uint64_t r;
>
> - asm volatile("vmcall"
> + if (is_amd_cpu())
> + asm volatile("vmmcall"
> : "=a"(r)
> : "a"(nr), "b"(a0), "c"(a1), "d"(a2), "S"(a3));
> + else
> + asm volatile("vmcall"
> + : "=a"(r)
> + : "a"(nr), "b"(a0), "c"(a1), "d"(a2), "S"(a3));
> +
> return r;
> }
>
> @@ -1314,8 +1321,10 @@ bool vm_is_unrestricted_guest(struct kvm_vm *vm)
>
> void kvm_arch_main(void)
> {
> + is_cpu_amd = cpu_vendor_string_is("AuthenticAMD") ? 1 : 0;
> }
>
> void kvm_arch_post_vm_load(struct kvm_vm *vm)
> {
> + sync_global_to_guest(vm, is_cpu_amd);
> }
> --
> 2.37.2.789.g6183377224-goog
>
On Thu, Sep 8, 2022 at 3:54 PM David Matlack <[email protected]> wrote:
>
> Please use "KVM: selftest: ..." for the shortlog.
>
Ack.
> On Sat, Sep 03, 2022 at 01:28:47AM +0000, Vishal Annapurve wrote:
> > Modify following APIs for x86 implementation:
> > 1) kvm_arch_main : Query the cpu vendor and cache the value in advance.
> > 2) kvm_arch_post_vm_load: Populate cpu type in the guest memory so that
> > guest doesn't need to execute cpuid instruction again.
>
> This commit message only describes a subset of the changes in this
> commit, and does not provide any context on why the changes are being
> made (other than a clue about avoiding CPUID).
>
> I also think this could be split up into 2 separate commits.
>
Ack, will add more details to the commit log. Though since
is_amd_cpu/is_intel_cpu is used by both selftest VMM and selftest
guest logic, I would prefer to not split this change into two. For
now, I will upload the series with this change being a single patch.
We can discuss again if this split is really needed.
Regards,
Vishal
> I would suggest first patch changes is_{intel,amd}_cpu() to return a cached
> result. e.g.
>
> KVM: selftests: Precompute the result for is_{intel,amd}_cpu()
>
> Cache the vendor CPU type in a global variable so that multiple calls
> to is_intel_cpu() do not need to re-execute CPUID. This will be useful
> in a follow-up commit to check if running on AMD or Intel from within
> a selftest guest where executing CPUID requires a VM-exit.
>
> Then add support for AMD to kvm_hypercall():
>
> KVM: selftests: Add AMD support to kvm_hypercall()
>
> Add support for AMD hypercalls to kvm_hypercall() so that it can be
> used in selftests running on Intel or AMD hosts. This will be used in
> a follow up commit to ...
>
> As part of this change, sync the global variable is_cpu_amd into the
> guest so the guest can determine which hypercall instruction to use
> without needing to re-execute CPUID for every hypercall.
>