2013-03-19 15:30:42

by Paolo Bonzini

[permalink] [raw]
Subject: [PATCH] x86: correctly initialize the CS base on reset

The CS base was initialized to 0 on VMX (wrong, but usually overridden
by userspace before starting) or 0xf0000 on SVM. The correct value is
0xffff0000, and VMX is able to emulate it now, so use it.

Signed-off-by: Paolo Bonzini <[email protected]>
---
arch/x86/kvm/svm.c | 8 +-------
arch/x86/kvm/vmx.c | 1 +
2 files changed, 2 insertions(+), 7 deletions(-)

diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 7219a40..7a46c1f 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1131,17 +1131,11 @@ static void init_vmcb(struct vcpu_svm *svm)
init_seg(&save->gs);

save->cs.selector = 0xf000;
+ save->cs.base = 0xffff0000;
/* Executable/Readable Code Segment */
save->cs.attrib = SVM_SELECTOR_READ_MASK | SVM_SELECTOR_P_MASK |
SVM_SELECTOR_S_MASK | SVM_SELECTOR_CODE_MASK;
save->cs.limit = 0xffff;
- /*
- * cs.base should really be 0xffff0000, but vmx can't handle that, so
- * be consistent with it.
- *
- * Replace when we have real mode working for vmx.
- */
- save->cs.base = 0xf0000;

save->gdtr.limit = 0xffff;
save->idtr.limit = 0xffff;
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 4a0bafe..c75c25d 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -4133,6 +4133,7 @@ static void vmx_vcpu_reset(struct kvm_vcpu *vcpu)

seg_setup(VCPU_SREG_CS);
vmcs_write16(GUEST_CS_SELECTOR, 0xf000);
+ vmcs_write32(GUEST_CS_BASE, 0xffff0000);

seg_setup(VCPU_SREG_DS);
seg_setup(VCPU_SREG_ES);
--
1.8.1.4


2013-03-19 15:43:25

by Gleb Natapov

[permalink] [raw]
Subject: Re: [PATCH] x86: correctly initialize the CS base on reset

On Tue, Mar 19, 2013 at 04:30:26PM +0100, Paolo Bonzini wrote:
> The CS base was initialized to 0 on VMX (wrong, but usually overridden
> by userspace before starting) or 0xf0000 on SVM. The correct value is
> 0xffff0000, and VMX is able to emulate it now, so use it.
>
> Signed-off-by: Paolo Bonzini <[email protected]>
Reviewed-by: Gleb Natapov <[email protected]>

> ---
> arch/x86/kvm/svm.c | 8 +-------
> arch/x86/kvm/vmx.c | 1 +
> 2 files changed, 2 insertions(+), 7 deletions(-)
>
> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
> index 7219a40..7a46c1f 100644
> --- a/arch/x86/kvm/svm.c
> +++ b/arch/x86/kvm/svm.c
> @@ -1131,17 +1131,11 @@ static void init_vmcb(struct vcpu_svm *svm)
> init_seg(&save->gs);
>
> save->cs.selector = 0xf000;
> + save->cs.base = 0xffff0000;
> /* Executable/Readable Code Segment */
> save->cs.attrib = SVM_SELECTOR_READ_MASK | SVM_SELECTOR_P_MASK |
> SVM_SELECTOR_S_MASK | SVM_SELECTOR_CODE_MASK;
> save->cs.limit = 0xffff;
> - /*
> - * cs.base should really be 0xffff0000, but vmx can't handle that, so
> - * be consistent with it.
> - *
> - * Replace when we have real mode working for vmx.
> - */
> - save->cs.base = 0xf0000;
>
> save->gdtr.limit = 0xffff;
> save->idtr.limit = 0xffff;
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 4a0bafe..c75c25d 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -4133,6 +4133,7 @@ static void vmx_vcpu_reset(struct kvm_vcpu *vcpu)
>
> seg_setup(VCPU_SREG_CS);
> vmcs_write16(GUEST_CS_SELECTOR, 0xf000);
> + vmcs_write32(GUEST_CS_BASE, 0xffff0000);
>
> seg_setup(VCPU_SREG_DS);
> seg_setup(VCPU_SREG_ES);
> --
> 1.8.1.4
>
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html

--
Gleb.

2013-03-19 16:41:52

by Jan Kiszka

[permalink] [raw]
Subject: Re: [PATCH] x86: correctly initialize the CS base on reset

On 2013-03-19 16:43, Gleb Natapov wrote:
> On Tue, Mar 19, 2013 at 04:30:26PM +0100, Paolo Bonzini wrote:
>> The CS base was initialized to 0 on VMX (wrong, but usually overridden
>> by userspace before starting) or 0xf0000 on SVM. The correct value is
>> 0xffff0000, and VMX is able to emulate it now, so use it.
>>
>> Signed-off-by: Paolo Bonzini <[email protected]>
> Reviewed-by: Gleb Natapov <[email protected]>

Just for the history (as I was wondering where this came from): CS base
used to be set to 0xf0000 on VMX as well, but that was changed by
b246dd5d to only affect guests when unrestricted mode is missing. That
change actually left the base uninitialized.

Jan

>
>> ---
>> arch/x86/kvm/svm.c | 8 +-------
>> arch/x86/kvm/vmx.c | 1 +
>> 2 files changed, 2 insertions(+), 7 deletions(-)
>>
>> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
>> index 7219a40..7a46c1f 100644
>> --- a/arch/x86/kvm/svm.c
>> +++ b/arch/x86/kvm/svm.c
>> @@ -1131,17 +1131,11 @@ static void init_vmcb(struct vcpu_svm *svm)
>> init_seg(&save->gs);
>>
>> save->cs.selector = 0xf000;
>> + save->cs.base = 0xffff0000;
>> /* Executable/Readable Code Segment */
>> save->cs.attrib = SVM_SELECTOR_READ_MASK | SVM_SELECTOR_P_MASK |
>> SVM_SELECTOR_S_MASK | SVM_SELECTOR_CODE_MASK;
>> save->cs.limit = 0xffff;
>> - /*
>> - * cs.base should really be 0xffff0000, but vmx can't handle that, so
>> - * be consistent with it.
>> - *
>> - * Replace when we have real mode working for vmx.
>> - */
>> - save->cs.base = 0xf0000;
>>
>> save->gdtr.limit = 0xffff;
>> save->idtr.limit = 0xffff;
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> index 4a0bafe..c75c25d 100644
>> --- a/arch/x86/kvm/vmx.c
>> +++ b/arch/x86/kvm/vmx.c
>> @@ -4133,6 +4133,7 @@ static void vmx_vcpu_reset(struct kvm_vcpu *vcpu)
>>
>> seg_setup(VCPU_SREG_CS);
>> vmcs_write16(GUEST_CS_SELECTOR, 0xf000);
>> + vmcs_write32(GUEST_CS_BASE, 0xffff0000);
>>
>> seg_setup(VCPU_SREG_DS);
>> seg_setup(VCPU_SREG_ES);
>> --
>> 1.8.1.4
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe kvm" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
> --
> Gleb.
>
--
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux

2013-03-19 17:08:00

by Gleb Natapov

[permalink] [raw]
Subject: Re: [PATCH] x86: correctly initialize the CS base on reset

On Tue, Mar 19, 2013 at 05:41:45PM +0100, Jan Kiszka wrote:
> On 2013-03-19 16:43, Gleb Natapov wrote:
> > On Tue, Mar 19, 2013 at 04:30:26PM +0100, Paolo Bonzini wrote:
> >> The CS base was initialized to 0 on VMX (wrong, but usually overridden
> >> by userspace before starting) or 0xf0000 on SVM. The correct value is
> >> 0xffff0000, and VMX is able to emulate it now, so use it.
> >>
> >> Signed-off-by: Paolo Bonzini <[email protected]>
> > Reviewed-by: Gleb Natapov <[email protected]>
>
> Just for the history (as I was wondering where this came from): CS base
> used to be set to 0xf0000 on VMX as well, but that was changed by
> b246dd5d to only affect guests when unrestricted mode is missing. That
> change actually left the base uninitialized.
>
Looking at the code before b246dd5d it sets CS base to 0xf0000 only if it was
0xffff0000. This is the same as with current code: fix_rmode_seg() does
it.

--
Gleb.

2013-03-20 23:19:19

by Marcelo Tosatti

[permalink] [raw]
Subject: Re: [PATCH] x86: correctly initialize the CS base on reset

On Tue, Mar 19, 2013 at 04:30:26PM +0100, Paolo Bonzini wrote:
> The CS base was initialized to 0 on VMX (wrong, but usually overridden
> by userspace before starting) or 0xf0000 on SVM. The correct value is
> 0xffff0000, and VMX is able to emulate it now, so use it.
>
> Signed-off-by: Paolo Bonzini <[email protected]>
> ---
> arch/x86/kvm/svm.c | 8 +-------
> arch/x86/kvm/vmx.c | 1 +
> 2 files changed, 2 insertions(+), 7 deletions(-)

Applied, thanks.