The scaling information in subleaf 1 should match the values set by KVM in
the 'vcpu_info' sub-structure 'time_info' (a.k.a. pvclock_vcpu_time_info)
which is shared with the guest, but is not directly available to the VMM.
The offset values are not set since a TSC offset is already applied.
The TSC frequency should also be set in sub-leaf 2.
Signed-off-by: Paul Durrant <[email protected]>
---
Cc: David Woodhouse <[email protected]>
v5:
- Drop the caching of the CPUID entry pointers and only update the
sub-leaves if the CPU frequency has actually changed
v4:
- Update commit comment
v3:
- Add leaf limit check in kvm_xen_set_cpuid()
v2:
- Make sure sub-leaf pointers are NULLed if the time leaf is removed
---
arch/x86/include/asm/kvm_host.h | 1 +
arch/x86/kvm/cpuid.c | 2 ++
arch/x86/kvm/x86.c | 1 +
arch/x86/kvm/xen.c | 51 +++++++++++++++++++++++++++++++++
arch/x86/kvm/xen.h | 10 +++++++
5 files changed, 65 insertions(+)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 88a3026ee163..abb0a39f60eb 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -638,6 +638,7 @@ struct kvm_vcpu_xen {
struct hrtimer timer;
int poll_evtchn;
struct timer_list poll_timer;
+ u32 cpuid_tsc_info;
};
struct kvm_vcpu_arch {
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index d47222ab8e6e..544d0f823ee5 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -25,6 +25,7 @@
#include "mmu.h"
#include "trace.h"
#include "pmu.h"
+#include "xen.h"
/*
* Unlike "struct cpuinfo_x86.x86_capability", kvm_cpu_caps doesn't need to be
@@ -310,6 +311,7 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
__cr4_reserved_bits(guest_cpuid_has, vcpu);
kvm_hv_set_cpuid(vcpu);
+ kvm_xen_after_set_cpuid(vcpu);
/* Invoke the vendor callback only after the above state is updated. */
static_call(kvm_x86_vcpu_after_set_cpuid)(vcpu);
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 031678eff28e..29ed665c51db 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -3110,6 +3110,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
&vcpu->hv_clock.tsc_shift,
&vcpu->hv_clock.tsc_to_system_mul);
vcpu->hw_tsc_khz = tgt_tsc_khz;
+ kvm_xen_setup_tsc_info(v);
}
vcpu->hv_clock.tsc_timestamp = tsc_timestamp;
diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c
index 610beba35907..c84424d5c8b6 100644
--- a/arch/x86/kvm/xen.c
+++ b/arch/x86/kvm/xen.c
@@ -10,6 +10,9 @@
#include "xen.h"
#include "hyperv.h"
#include "lapic.h"
+#include "cpuid.h"
+
+#include <asm/xen/cpuid.h>
#include <linux/eventfd.h>
#include <linux/kvm_host.h>
@@ -1855,3 +1858,51 @@ void kvm_xen_destroy_vm(struct kvm *kvm)
if (kvm->arch.xen_hvm_config.msr)
static_branch_slow_dec_deferred(&kvm_xen_enabled);
}
+
+void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
+{
+ u32 base = 0;
+ u32 limit;
+ u32 function;
+
+ vcpu->arch.xen.cpuid_tsc_info = 0;
+
+ for_each_possible_hypervisor_cpuid_base(function) {
+ struct kvm_cpuid_entry2 *entry = kvm_find_cpuid_entry(vcpu, function, 0);
+
+ if (entry &&
+ entry->ebx == XEN_CPUID_SIGNATURE_EBX &&
+ entry->ecx == XEN_CPUID_SIGNATURE_ECX &&
+ entry->edx == XEN_CPUID_SIGNATURE_EDX) {
+ base = function;
+ limit = entry->eax;
+ break;
+ }
+ }
+ if (!base)
+ return;
+
+ function = base | XEN_CPUID_LEAF(3);
+ if (function > limit)
+ return;
+
+ vcpu->arch.xen.cpuid_tsc_info = function;
+}
+
+void kvm_xen_setup_tsc_info(struct kvm_vcpu *vcpu)
+{
+ struct kvm_cpuid_entry2 *entry;
+
+ if (!vcpu->arch.xen.cpuid_tsc_info)
+ return;
+
+ entry = kvm_find_cpuid_entry(vcpu, vcpu->arch.xen.cpuid_tsc_info, 1);
+ if (entry) {
+ entry->ecx = vcpu->arch.hv_clock.tsc_to_system_mul;
+ entry->edx = vcpu->arch.hv_clock.tsc_shift;
+ }
+
+ entry = kvm_find_cpuid_entry(vcpu, vcpu->arch.xen.cpuid_tsc_info, 2);
+ if (entry)
+ entry->eax = vcpu->arch.hw_tsc_khz;
+}
diff --git a/arch/x86/kvm/xen.h b/arch/x86/kvm/xen.h
index 532a535a9e99..b2ca434431d6 100644
--- a/arch/x86/kvm/xen.h
+++ b/arch/x86/kvm/xen.h
@@ -32,6 +32,8 @@ int kvm_xen_set_evtchn_fast(struct kvm_xen_evtchn *xe,
int kvm_xen_setup_evtchn(struct kvm *kvm,
struct kvm_kernel_irq_routing_entry *e,
const struct kvm_irq_routing_entry *ue);
+void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu);
+void kvm_xen_setup_tsc_info(struct kvm_vcpu *vcpu);
static inline bool kvm_xen_msr_enabled(struct kvm *kvm)
{
@@ -135,6 +137,14 @@ static inline bool kvm_xen_timer_enabled(struct kvm_vcpu *vcpu)
{
return false;
}
+
+static inline void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
+{
+}
+
+static inline void kvm_xen_setup_tsc_info(struct kvm_vcpu *vcpu)
+{
+}
#endif
int kvm_xen_hypercall(struct kvm_vcpu *vcpu);
--
2.20.1
On Wed, Jun 29, 2022, Paul Durrant wrote:
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 88a3026ee163..abb0a39f60eb 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -638,6 +638,7 @@ struct kvm_vcpu_xen {
> struct hrtimer timer;
> int poll_evtchn;
> struct timer_list poll_timer;
> + u32 cpuid_tsc_info;
I would prefer to follow vcpu->arch.kvm_cpuid_base and capture the base CPUID
function. I have a hard time believing this will be the only case where KVM needs
to query XEN CPUID leafs. And cpuid_tsc_info is a confusing name given the helper
kvm_xen_setup_tsc_info(); it's odd to see a "setup" helper immediately consume a
variable with the same name.
It'll incur another CPUID lookup in the update path to check the limit, but again
that should be a rare operation so it doesn't seem too onerous.
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 031678eff28e..29ed665c51db 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -3110,6 +3110,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
> &vcpu->hv_clock.tsc_shift,
> &vcpu->hv_clock.tsc_to_system_mul);
> vcpu->hw_tsc_khz = tgt_tsc_khz;
> + kvm_xen_setup_tsc_info(v);
Any objection to s/setup/update? KVM Xen uses "setup" for things like configuring
the event channel using userspace input, whereas this is purely updating existing
data structures.
> }
>
> vcpu->hv_clock.tsc_timestamp = tsc_timestamp;
> diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c
> index 610beba35907..c84424d5c8b6 100644
> --- a/arch/x86/kvm/xen.c
> +++ b/arch/x86/kvm/xen.c
> @@ -10,6 +10,9 @@
> #include "xen.h"
> #include "hyperv.h"
> #include "lapic.h"
> +#include "cpuid.h"
> +
> +#include <asm/xen/cpuid.h>
>
> #include <linux/eventfd.h>
> #include <linux/kvm_host.h>
> @@ -1855,3 +1858,51 @@ void kvm_xen_destroy_vm(struct kvm *kvm)
> if (kvm->arch.xen_hvm_config.msr)
> static_branch_slow_dec_deferred(&kvm_xen_enabled);
> }
> +
> +void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
> +{
> + u32 base = 0;
> + u32 limit;
> + u32 function;
> +
> + vcpu->arch.xen.cpuid_tsc_info = 0;
> +
> + for_each_possible_hypervisor_cpuid_base(function) {
> + struct kvm_cpuid_entry2 *entry = kvm_find_cpuid_entry(vcpu, function, 0);
> +
> + if (entry &&
> + entry->ebx == XEN_CPUID_SIGNATURE_EBX &&
> + entry->ecx == XEN_CPUID_SIGNATURE_ECX &&
> + entry->edx == XEN_CPUID_SIGNATURE_EDX) {
> + base = function;
> + limit = entry->eax;
> + break;
> + }
> + }
> + if (!base)
> + return;
Rather than open code a variant of kvm_update_kvm_cpuid_base(), that helper can
be tweaked to take a signature. Along with a patch to provide a #define for Xen's
signature as a string, this entire function becomes a one-liner.
If the below looks ok (won't compile, needs prep patches), I'll test and post a
proper mini-series.
---
arch/x86/include/asm/kvm_host.h | 1 +
arch/x86/kvm/cpuid.c | 2 ++
arch/x86/kvm/x86.c | 1 +
arch/x86/kvm/xen.c | 30 ++++++++++++++++++++++++++++++
arch/x86/kvm/xen.h | 22 +++++++++++++++++++++-
5 files changed, 55 insertions(+), 1 deletion(-)
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index de5a149d0971..b2565d05fc86 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -638,6 +638,7 @@ struct kvm_vcpu_xen {
struct hrtimer timer;
int poll_evtchn;
struct timer_list poll_timer;
+ u32 cpuid_base;
};
struct kvm_vcpu_arch {
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index 0abe3adc9ae3..54ed51799b8d 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -25,6 +25,7 @@
#include "mmu.h"
#include "trace.h"
#include "pmu.h"
+#include "xen.h"
/*
* Unlike "struct cpuinfo_x86.x86_capability", kvm_cpu_caps doesn't need to be
@@ -309,6 +310,7 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
__cr4_reserved_bits(guest_cpuid_has, vcpu);
kvm_hv_set_cpuid(vcpu);
+ kvm_xen_after_set_cpuid(vcpu);
/* Invoke the vendor callback only after the above state is updated. */
static_call(kvm_x86_vcpu_after_set_cpuid)(vcpu);
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 567d13405445..a624293c66c8 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -3110,6 +3110,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
&vcpu->hv_clock.tsc_shift,
&vcpu->hv_clock.tsc_to_system_mul);
vcpu->hw_tsc_khz = tgt_tsc_khz;
+ kvm_xen_update_tsc_info(v);
}
vcpu->hv_clock.tsc_timestamp = tsc_timestamp;
diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c
index 610beba35907..3fc0c194b813 100644
--- a/arch/x86/kvm/xen.c
+++ b/arch/x86/kvm/xen.c
@@ -10,6 +10,9 @@
#include "xen.h"
#include "hyperv.h"
#include "lapic.h"
+#include "cpuid.h"
+
+#include <asm/xen/cpuid.h>
#include <linux/eventfd.h>
#include <linux/kvm_host.h>
@@ -1855,3 +1858,30 @@ void kvm_xen_destroy_vm(struct kvm *kvm)
if (kvm->arch.xen_hvm_config.msr)
static_branch_slow_dec_deferred(&kvm_xen_enabled);
}
+
+void kvm_xen_update_tsc_info(struct kvm_vcpu *vcpu)
+{
+ struct kvm_cpuid_entry2 *entry;
+ u32 function;
+
+ if (!vcpu->arch.xen.cpuid_base)
+ return;
+
+ entry = kvm_find_cpuid_entry(vcpu, vcpu->arch.xen.cpuid_base, 0);
+ if (WARN_ON_ONCE(!entry))
+ return;
+
+ function = vcpu->arch.xen.cpuid_base | XEN_CPUID_LEAF(3);
+ if (function > entry->eax)
+ return;
+
+ entry = kvm_find_cpuid_entry(vcpu, function, 1);
+ if (entry) {
+ entry->ecx = vcpu->arch.hv_clock.tsc_to_system_mul;
+ entry->edx = vcpu->arch.hv_clock.tsc_shift;
+ }
+
+ entry = kvm_find_cpuid_entry(vcpu, function, 2);
+ if (entry)
+ entry->eax = vcpu->arch.hw_tsc_khz;
+}
diff --git a/arch/x86/kvm/xen.h b/arch/x86/kvm/xen.h
index 532a535a9e99..b8161b99b82a 100644
--- a/arch/x86/kvm/xen.h
+++ b/arch/x86/kvm/xen.h
@@ -9,9 +9,14 @@
#ifndef __ARCH_X86_KVM_XEN_H__
#define __ARCH_X86_KVM_XEN_H__
-#ifdef CONFIG_KVM_XEN
#include <linux/jump_label_ratelimit.h>
+#include <asm/xen/cpuid.h>
+
+#include "cpuid.h"
+
+#ifdef CONFIG_KVM_XEN
+
extern struct static_key_false_deferred kvm_xen_enabled;
int __kvm_xen_has_interrupt(struct kvm_vcpu *vcpu);
@@ -32,6 +37,13 @@ int kvm_xen_set_evtchn_fast(struct kvm_xen_evtchn *xe,
int kvm_xen_setup_evtchn(struct kvm *kvm,
struct kvm_kernel_irq_routing_entry *e,
const struct kvm_irq_routing_entry *ue);
+void kvm_xen_update_tsc_info(struct kvm_vcpu *vcpu);
+
+static inline void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
+{
+ vcpu->arch.xen.cpuid_base =
+ kvm_get_hypervisor_cpuid_base(vcpu, XEN_CPUID_SIGNATURE);
+}
static inline bool kvm_xen_msr_enabled(struct kvm *kvm)
{
@@ -135,6 +147,14 @@ static inline bool kvm_xen_timer_enabled(struct kvm_vcpu *vcpu)
{
return false;
}
+
+static inline void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
+{
+}
+
+static inline void kvm_xen_update_tsc_info(struct kvm_vcpu *vcpu)
+{
+}
#endif
int kvm_xen_hypercall(struct kvm_vcpu *vcpu);
base-commit: b08b2f54c49d8f96a22107c444d500dff73ec2a6
--
> -----Original Message-----
> From: Sean Christopherson <[email protected]>
> Sent: 12 July 2022 00:44
> To: Durrant, Paul <[email protected]>
> Cc: [email protected]; [email protected]; [email protected]; David Woodhouse
> <[email protected]>; Paolo Bonzini <[email protected]>; Vitaly Kuznetsov <[email protected]>;
> Wanpeng Li <[email protected]>; Jim Mattson <[email protected]>; Joerg Roedel <[email protected]>;
> Thomas Gleixner <[email protected]>; Ingo Molnar <[email protected]>; Borislav Petkov <[email protected]>;
> Dave Hansen <[email protected]>; H. Peter Anvin <[email protected]>
> Subject: RE: [EXTERNAL][PATCH v5] KVM: x86/xen: Update Xen CPUID Leaf 4 (tsc info) sub-leaves, if
> present
>
> CAUTION: This email originated from outside of the organization. Do not click links or open
> attachments unless you can confirm the sender and know the content is safe.
>
>
>
> On Wed, Jun 29, 2022, Paul Durrant wrote:
> > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> > index 88a3026ee163..abb0a39f60eb 100644
> > --- a/arch/x86/include/asm/kvm_host.h
> > +++ b/arch/x86/include/asm/kvm_host.h
> > @@ -638,6 +638,7 @@ struct kvm_vcpu_xen {
> > struct hrtimer timer;
> > int poll_evtchn;
> > struct timer_list poll_timer;
> > + u32 cpuid_tsc_info;
>
> I would prefer to follow vcpu->arch.kvm_cpuid_base and capture the base CPUID
> function. I have a hard time believing this will be the only case where KVM needs
> to query XEN CPUID leafs. And cpuid_tsc_info is a confusing name given the helper
> kvm_xen_setup_tsc_info(); it's odd to see a "setup" helper immediately consume a
> variable with the same name.
Sure. It is rather shrink-to-fit at the moment... no problem with capturing the base.
>
> It'll incur another CPUID lookup in the update path to check the limit, but again
> that should be a rare operation so it doesn't seem too onerous.
>
We could capture the limit leaf in the general case. It's not Xen-specific after all.
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index 031678eff28e..29ed665c51db 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -3110,6 +3110,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
> > &vcpu->hv_clock.tsc_shift,
> > &vcpu->hv_clock.tsc_to_system_mul);
> > vcpu->hw_tsc_khz = tgt_tsc_khz;
> > + kvm_xen_setup_tsc_info(v);
>
> Any objection to s/setup/update? KVM Xen uses "setup" for things like configuring
> the event channel using userspace input, whereas this is purely updating existing
> data structures.
>
Sure.
> > }
> >
> > vcpu->hv_clock.tsc_timestamp = tsc_timestamp;
> > diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c
> > index 610beba35907..c84424d5c8b6 100644
> > --- a/arch/x86/kvm/xen.c
> > +++ b/arch/x86/kvm/xen.c
> > @@ -10,6 +10,9 @@
> > #include "xen.h"
> > #include "hyperv.h"
> > #include "lapic.h"
> > +#include "cpuid.h"
> > +
> > +#include <asm/xen/cpuid.h>
> >
> > #include <linux/eventfd.h>
> > #include <linux/kvm_host.h>
> > @@ -1855,3 +1858,51 @@ void kvm_xen_destroy_vm(struct kvm *kvm)
> > if (kvm->arch.xen_hvm_config.msr)
> > static_branch_slow_dec_deferred(&kvm_xen_enabled);
> > }
> > +
> > +void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
> > +{
> > + u32 base = 0;
> > + u32 limit;
> > + u32 function;
> > +
> > + vcpu->arch.xen.cpuid_tsc_info = 0;
> > +
> > + for_each_possible_hypervisor_cpuid_base(function) {
> > + struct kvm_cpuid_entry2 *entry = kvm_find_cpuid_entry(vcpu, function, 0);
> > +
> > + if (entry &&
> > + entry->ebx == XEN_CPUID_SIGNATURE_EBX &&
> > + entry->ecx == XEN_CPUID_SIGNATURE_ECX &&
> > + entry->edx == XEN_CPUID_SIGNATURE_EDX) {
> > + base = function;
> > + limit = entry->eax;
> > + break;
> > + }
> > + }
> > + if (!base)
> > + return;
>
> Rather than open code a variant of kvm_update_kvm_cpuid_base(), that helper can
> be tweaked to take a signature. Along with a patch to provide a #define for Xen's
> signature as a string, this entire function becomes a one-liner.
>
Sure, but as said above, we could make capturing the limit part of the general function too. It could even be extended to capture the Hyper-V base/limit too.
As for defining the sig as a string... I guess it would be neater to use the values from the Xen header, but it'll probably make the code more ugly so a secondary definition is reasonable.
> If the below looks ok (won't compile, needs prep patches), I'll test and post a
> proper mini-series.
Ok. Thanks,
Paul
>
> ---
> arch/x86/include/asm/kvm_host.h | 1 +
> arch/x86/kvm/cpuid.c | 2 ++
> arch/x86/kvm/x86.c | 1 +
> arch/x86/kvm/xen.c | 30 ++++++++++++++++++++++++++++++
> arch/x86/kvm/xen.h | 22 +++++++++++++++++++++-
> 5 files changed, 55 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index de5a149d0971..b2565d05fc86 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -638,6 +638,7 @@ struct kvm_vcpu_xen {
> struct hrtimer timer;
> int poll_evtchn;
> struct timer_list poll_timer;
> + u32 cpuid_base;
> };
>
> struct kvm_vcpu_arch {
> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
> index 0abe3adc9ae3..54ed51799b8d 100644
> --- a/arch/x86/kvm/cpuid.c
> +++ b/arch/x86/kvm/cpuid.c
> @@ -25,6 +25,7 @@
> #include "mmu.h"
> #include "trace.h"
> #include "pmu.h"
> +#include "xen.h"
>
> /*
> * Unlike "struct cpuinfo_x86.x86_capability", kvm_cpu_caps doesn't need to be
> @@ -309,6 +310,7 @@ static void kvm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
> __cr4_reserved_bits(guest_cpuid_has, vcpu);
>
> kvm_hv_set_cpuid(vcpu);
> + kvm_xen_after_set_cpuid(vcpu);
>
> /* Invoke the vendor callback only after the above state is updated. */
> static_call(kvm_x86_vcpu_after_set_cpuid)(vcpu);
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 567d13405445..a624293c66c8 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -3110,6 +3110,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v)
> &vcpu->hv_clock.tsc_shift,
> &vcpu->hv_clock.tsc_to_system_mul);
> vcpu->hw_tsc_khz = tgt_tsc_khz;
> + kvm_xen_update_tsc_info(v);
> }
>
> vcpu->hv_clock.tsc_timestamp = tsc_timestamp;
> diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c
> index 610beba35907..3fc0c194b813 100644
> --- a/arch/x86/kvm/xen.c
> +++ b/arch/x86/kvm/xen.c
> @@ -10,6 +10,9 @@
> #include "xen.h"
> #include "hyperv.h"
> #include "lapic.h"
> +#include "cpuid.h"
> +
> +#include <asm/xen/cpuid.h>
>
> #include <linux/eventfd.h>
> #include <linux/kvm_host.h>
> @@ -1855,3 +1858,30 @@ void kvm_xen_destroy_vm(struct kvm *kvm)
> if (kvm->arch.xen_hvm_config.msr)
> static_branch_slow_dec_deferred(&kvm_xen_enabled);
> }
> +
> +void kvm_xen_update_tsc_info(struct kvm_vcpu *vcpu)
> +{
> + struct kvm_cpuid_entry2 *entry;
> + u32 function;
> +
> + if (!vcpu->arch.xen.cpuid_base)
> + return;
> +
> + entry = kvm_find_cpuid_entry(vcpu, vcpu->arch.xen.cpuid_base, 0);
> + if (WARN_ON_ONCE(!entry))
> + return;
> +
> + function = vcpu->arch.xen.cpuid_base | XEN_CPUID_LEAF(3);
> + if (function > entry->eax)
> + return;
> +
> + entry = kvm_find_cpuid_entry(vcpu, function, 1);
> + if (entry) {
> + entry->ecx = vcpu->arch.hv_clock.tsc_to_system_mul;
> + entry->edx = vcpu->arch.hv_clock.tsc_shift;
> + }
> +
> + entry = kvm_find_cpuid_entry(vcpu, function, 2);
> + if (entry)
> + entry->eax = vcpu->arch.hw_tsc_khz;
> +}
> diff --git a/arch/x86/kvm/xen.h b/arch/x86/kvm/xen.h
> index 532a535a9e99..b8161b99b82a 100644
> --- a/arch/x86/kvm/xen.h
> +++ b/arch/x86/kvm/xen.h
> @@ -9,9 +9,14 @@
> #ifndef __ARCH_X86_KVM_XEN_H__
> #define __ARCH_X86_KVM_XEN_H__
>
> -#ifdef CONFIG_KVM_XEN
> #include <linux/jump_label_ratelimit.h>
>
> +#include <asm/xen/cpuid.h>
> +
> +#include "cpuid.h"
> +
> +#ifdef CONFIG_KVM_XEN
> +
> extern struct static_key_false_deferred kvm_xen_enabled;
>
> int __kvm_xen_has_interrupt(struct kvm_vcpu *vcpu);
> @@ -32,6 +37,13 @@ int kvm_xen_set_evtchn_fast(struct kvm_xen_evtchn *xe,
> int kvm_xen_setup_evtchn(struct kvm *kvm,
> struct kvm_kernel_irq_routing_entry *e,
> const struct kvm_irq_routing_entry *ue);
> +void kvm_xen_update_tsc_info(struct kvm_vcpu *vcpu);
> +
> +static inline void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
> +{
> + vcpu->arch.xen.cpuid_base =
> + kvm_get_hypervisor_cpuid_base(vcpu, XEN_CPUID_SIGNATURE);
> +}
>
> static inline bool kvm_xen_msr_enabled(struct kvm *kvm)
> {
> @@ -135,6 +147,14 @@ static inline bool kvm_xen_timer_enabled(struct kvm_vcpu *vcpu)
> {
> return false;
> }
> +
> +static inline void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
> +{
> +}
> +
> +static inline void kvm_xen_update_tsc_info(struct kvm_vcpu *vcpu)
> +{
> +}
> #endif
>
> int kvm_xen_hypercall(struct kvm_vcpu *vcpu);
>
> base-commit: b08b2f54c49d8f96a22107c444d500dff73ec2a6
> --
On Tue, Jul 12, 2022, Durrant, Paul wrote:
> > > @@ -1855,3 +1858,51 @@ void kvm_xen_destroy_vm(struct kvm *kvm)
> > > if (kvm->arch.xen_hvm_config.msr)
> > > static_branch_slow_dec_deferred(&kvm_xen_enabled);
> > > }
> > > +
> > > +void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
> > > +{
> > > + u32 base = 0;
> > > + u32 limit;
> > > + u32 function;
> > > +
> > > + vcpu->arch.xen.cpuid_tsc_info = 0;
> > > +
> > > + for_each_possible_hypervisor_cpuid_base(function) {
> > > + struct kvm_cpuid_entry2 *entry = kvm_find_cpuid_entry(vcpu, function, 0);
> > > +
> > > + if (entry &&
> > > + entry->ebx == XEN_CPUID_SIGNATURE_EBX &&
> > > + entry->ecx == XEN_CPUID_SIGNATURE_ECX &&
> > > + entry->edx == XEN_CPUID_SIGNATURE_EDX) {
> > > + base = function;
> > > + limit = entry->eax;
> > > + break;
> > > + }
> > > + }
> > > + if (!base)
> > > + return;
> >
> > Rather than open code a variant of kvm_update_kvm_cpuid_base(), that helper can
> > be tweaked to take a signature. Along with a patch to provide a #define for Xen's
> > signature as a string, this entire function becomes a one-liner.
> >
>
> Sure, but as said above, we could make capturing the limit part of the
> general function too. It could even be extended to capture the Hyper-V
> base/limit too. As for defining the sig as a string... I guess it would be
> neater to use the values from the Xen header, but it'll probably make the
> code more ugly so a secondary definition is reasonable.
The base needs to be captured separately for KVM and Xen because KVM (and presumably
Xen itself since Xen also allows a variable base) supports advertising multiple
hypervisors to the guest. I don't know if there are any guests that will concurrently
utilize multiple hypervisor's paravirt features, so maybe we could squeak by, but
saving 4 bytes isn't worth the risk.
AFAIK, Hyper-V doesn't allow for a variable base, and so doesn't utilize the
for_each_possible... macro.
Sean Christopherson <[email protected]> writes:
> On Tue, Jul 12, 2022, Durrant, Paul wrote:
>> > > @@ -1855,3 +1858,51 @@ void kvm_xen_destroy_vm(struct kvm *kvm)
>> > > if (kvm->arch.xen_hvm_config.msr)
>> > > static_branch_slow_dec_deferred(&kvm_xen_enabled);
>> > > }
>> > > +
>> > > +void kvm_xen_after_set_cpuid(struct kvm_vcpu *vcpu)
>> > > +{
>> > > + u32 base = 0;
>> > > + u32 limit;
>> > > + u32 function;
>> > > +
>> > > + vcpu->arch.xen.cpuid_tsc_info = 0;
>> > > +
>> > > + for_each_possible_hypervisor_cpuid_base(function) {
>> > > + struct kvm_cpuid_entry2 *entry = kvm_find_cpuid_entry(vcpu, function, 0);
>> > > +
>> > > + if (entry &&
>> > > + entry->ebx == XEN_CPUID_SIGNATURE_EBX &&
>> > > + entry->ecx == XEN_CPUID_SIGNATURE_ECX &&
>> > > + entry->edx == XEN_CPUID_SIGNATURE_EDX) {
>> > > + base = function;
>> > > + limit = entry->eax;
>> > > + break;
>> > > + }
>> > > + }
>> > > + if (!base)
>> > > + return;
>> >
>> > Rather than open code a variant of kvm_update_kvm_cpuid_base(), that helper can
>> > be tweaked to take a signature. Along with a patch to provide a #define for Xen's
>> > signature as a string, this entire function becomes a one-liner.
>> >
>>
>> Sure, but as said above, we could make capturing the limit part of the
>> general function too. It could even be extended to capture the Hyper-V
>> base/limit too. As for defining the sig as a string... I guess it would be
>> neater to use the values from the Xen header, but it'll probably make the
>> code more ugly so a secondary definition is reasonable.
>
> The base needs to be captured separately for KVM and Xen because KVM (and presumably
> Xen itself since Xen also allows a variable base) supports advertising multiple
> hypervisors to the guest. I don't know if there are any guests that will concurrently
> utilize multiple hypervisor's paravirt features, so maybe we could squeak by, but
> saving 4 bytes isn't worth the risk.
>
> AFAIK, Hyper-V doesn't allow for a variable base, and so doesn't utilize the
> for_each_possible... macro.
FWIW, this matches my understanding too: Windows guests don't seem to
check anything besides 0x40000001 and give up if Hyper-V's id is not
there.
--
Vitaly