[PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@redhat.com>
---
 arch/x86/kvm/x86.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index f6a31a1..e2b0e36 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq)
 
 	vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
 	vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
+	vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
 	ret = emulate_int_real(&vcpu->arch.emulate_ctxt, &emulate_ops, irq);
 
 	if (ret != X86EMUL_CONTINUE)
-- 
1.7.1

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Avi Kivity]

  On 08/18/2010 10:36 AM, Avi Kivity wrote:
> Signed-off-by: Avi Kivity<avi@redhat.com>
> ---
>   arch/x86/kvm/x86.c |    1 +
>   1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index f6a31a1..e2b0e36 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq)
>
>   	vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
>   	vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
> +	vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
>   	ret = emulate_int_real(&vcpu->arch.emulate_ctxt,&emulate_ops, irq);
>
>   	if (ret != X86EMUL_CONTINUE)

With this, the non-atomic-injection branch trips into an unemulated DAS 
instruction (when it doesn't inject an interrupt at the wrong place).

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Mohammed Gamal]

On Wed, Aug 18, 2010 at 10:36 AM, Avi Kivity <avi@redhat.com> wrote:
> Signed-off-by: Avi Kivity <avi@redhat.com>
> ---
>  arch/x86/kvm/x86.c |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index f6a31a1..e2b0e36 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq)
>
>        vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
>        vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
> +       vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
I think it'd make more sense to make
vcpu->arch.emulate_ctxt.decode.eip =  vcpu->arch.emulate_ctxt.eip, as
we already initlialize the emulation context eip. It's not going to
make any real difference, but it's more readable this way I suppose,
and is symmetric to the assignment of the emulation context eip value
after the call to emulate_int_real()

>        ret = emulate_int_real(&vcpu->arch.emulate_ctxt, &emulate_ops, irq);
>
>        if (ret != X86EMUL_CONTINUE)
> --
> 1.7.1
>
>

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Avi Kivity]

  On 08/18/2010 03:46 PM, Mohammed Gamal wrote:
> On Wed, Aug 18, 2010 at 10:36 AM, Avi Kivity<avi@redhat.com>  wrote:
>> Signed-off-by: Avi Kivity<avi@redhat.com>
>> ---
>>   arch/x86/kvm/x86.c |    1 +
>>   1 files changed, 1 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index f6a31a1..e2b0e36 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq)
>>
>>         vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
>>         vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
>> +       vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
> I think it'd make more sense to make
> vcpu->arch.emulate_ctxt.decode.eip =  vcpu->arch.emulate_ctxt.eip, as
> we already initlialize the emulation context eip. It's not going to
> make any real difference, but it's more readable this way I suppose,
> and is symmetric to the assignment of the emulation context eip value
> after the call to emulate_int_real()

Right, patch updated.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Avi Kivity]

  On 08/18/2010 05:09 PM, Avi Kivity wrote:
>  On 08/18/2010 03:46 PM, Mohammed Gamal wrote:
>> On Wed, Aug 18, 2010 at 10:36 AM, Avi Kivity<avi@redhat.com>  wrote:
>>> Signed-off-by: Avi Kivity<avi@redhat.com>
>>> ---
>>>   arch/x86/kvm/x86.c |    1 +
>>>   1 files changed, 1 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>>> index f6a31a1..e2b0e36 100644
>>> --- a/arch/x86/kvm/x86.c
>>> +++ b/arch/x86/kvm/x86.c
>>> @@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct 
>>> kvm_vcpu *vcpu, int irq)
>>>
>>>         vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
>>>         vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
>>> +       vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
>> I think it'd make more sense to make
>> vcpu->arch.emulate_ctxt.decode.eip =  vcpu->arch.emulate_ctxt.eip, as
>> we already initlialize the emulation context eip. It's not going to
>> make any real difference, but it's more readable this way I suppose,
>> and is symmetric to the assignment of the emulation context eip value
>> after the call to emulate_int_real()
>
> Right, patch updated.
>

btw, with the other fixes in non-atomic-injection the bios is able to 
boot  up to the "no bootable device" message.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Mohammed Gamal]

On Wed, Aug 18, 2010 at 5:11 PM, Avi Kivity <avi@redhat.com> wrote:
>  On 08/18/2010 05:09 PM, Avi Kivity wrote:
>>
>>  On 08/18/2010 03:46 PM, Mohammed Gamal wrote:
>>>
>>> On Wed, Aug 18, 2010 at 10:36 AM, Avi Kivity<avi@redhat.com>  wrote:
>>>>
>>>> Signed-off-by: Avi Kivity<avi@redhat.com>
>>>> ---
>>>>  arch/x86/kvm/x86.c |    1 +
>>>>  1 files changed, 1 insertions(+), 0 deletions(-)
>>>>
>>>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>>>> index f6a31a1..e2b0e36 100644
>>>> --- a/arch/x86/kvm/x86.c
>>>> +++ b/arch/x86/kvm/x86.c
>>>> @@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct kvm_vcpu
>>>> *vcpu, int irq)
>>>>
>>>>        vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
>>>>        vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
>>>> +       vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
>>>
>>> I think it'd make more sense to make
>>> vcpu->arch.emulate_ctxt.decode.eip =  vcpu->arch.emulate_ctxt.eip, as
>>> we already initlialize the emulation context eip. It's not going to
>>> make any real difference, but it's more readable this way I suppose,
>>> and is symmetric to the assignment of the emulation context eip value
>>> after the call to emulate_int_real()
>>
>> Right, patch updated.
>>
>
> btw, with the other fixes in non-atomic-injection the bios is able to boot
>  up to the "no bootable device" message.
Did u try running it with any actual guests?

>
> --
> I have a truly marvellous patch that fixes the bug which this
> signature is too narrow to contain.
>
>

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Avi Kivity]

  On 08/18/2010 05:32 PM, Mohammed Gamal wrote:
> On Wed, Aug 18, 2010 at 5:11 PM, Avi Kivity<avi@redhat.com>  wrote:
>>   On 08/18/2010 05:09 PM, Avi Kivity wrote:
>>>   On 08/18/2010 03:46 PM, Mohammed Gamal wrote:
>>>> On Wed, Aug 18, 2010 at 10:36 AM, Avi Kivity<avi@redhat.com>    wrote:
>>>>> Signed-off-by: Avi Kivity<avi@redhat.com>
>>>>> ---
>>>>>   arch/x86/kvm/x86.c |    1 +
>>>>>   1 files changed, 1 insertions(+), 0 deletions(-)
>>>>>
>>>>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>>>>> index f6a31a1..e2b0e36 100644
>>>>> --- a/arch/x86/kvm/x86.c
>>>>> +++ b/arch/x86/kvm/x86.c
>>>>> @@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct kvm_vcpu
>>>>> *vcpu, int irq)
>>>>>
>>>>>         vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
>>>>>         vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
>>>>> +       vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
>>>> I think it'd make more sense to make
>>>> vcpu->arch.emulate_ctxt.decode.eip =  vcpu->arch.emulate_ctxt.eip, as
>>>> we already initlialize the emulation context eip. It's not going to
>>>> make any real difference, but it's more readable this way I suppose,
>>>> and is symmetric to the assignment of the emulation context eip value
>>>> after the call to emulate_int_real()
>>> Right, patch updated.
>>>
>> btw, with the other fixes in non-atomic-injection the bios is able to boot
>>   up to the "no bootable device" message.
> Did u try running it with any actual guests?

I did, the disk is not recognized, so something's still broken.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Mohammed Gamal]

On Wed, Aug 18, 2010 at 5:46 PM, Avi Kivity <avi@redhat.com> wrote:
>  On 08/18/2010 05:32 PM, Mohammed Gamal wrote:
>>
>> On Wed, Aug 18, 2010 at 5:11 PM, Avi Kivity<avi@redhat.com>  wrote:
>>>
>>>  On 08/18/2010 05:09 PM, Avi Kivity wrote:
>>>>
>>>>  On 08/18/2010 03:46 PM, Mohammed Gamal wrote:
>>>>>
>>>>> On Wed, Aug 18, 2010 at 10:36 AM, Avi Kivity<avi@redhat.com>    wrote:
>>>>>>
>>>>>> Signed-off-by: Avi Kivity<avi@redhat.com>
>>>>>> ---
>>>>>>  arch/x86/kvm/x86.c |    1 +
>>>>>>  1 files changed, 1 insertions(+), 0 deletions(-)
>>>>>>
>>>>>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>>>>>> index f6a31a1..e2b0e36 100644
>>>>>> --- a/arch/x86/kvm/x86.c
>>>>>> +++ b/arch/x86/kvm/x86.c
>>>>>> @@ -3967,6 +3967,7 @@ int kvm_inject_realmode_interrupt(struct
>>>>>> kvm_vcpu
>>>>>> *vcpu, int irq)
>>>>>>
>>>>>>        vcpu->arch.emulate_ctxt.decode.op_bytes = 2;
>>>>>>        vcpu->arch.emulate_ctxt.decode.ad_bytes = 2;
>>>>>> +       vcpu->arch.emulate_ctxt.decode.eip = kvm_rip_read(vcpu);
>>>>>
>>>>> I think it'd make more sense to make
>>>>> vcpu->arch.emulate_ctxt.decode.eip =  vcpu->arch.emulate_ctxt.eip, as
>>>>> we already initlialize the emulation context eip. It's not going to
>>>>> make any real difference, but it's more readable this way I suppose,
>>>>> and is symmetric to the assignment of the emulation context eip value
>>>>> after the call to emulate_int_real()
>>>>
>>>> Right, patch updated.
>>>>
>>> btw, with the other fixes in non-atomic-injection the bios is able to
>>> boot
>>>  up to the "no bootable device" message.
>>
>> Did u try running it with any actual guests?
>
> I did, the disk is not recognized, so something's still broken.
So far I get hard disk images recognized, but cdrom images aren't.
>
> --
> I have a truly marvellous patch that fixes the bug which this
> signature is too narrow to contain.
>
>

Re: [PATCH non-atomic-injection] KVM: Initialize rip for real mode interrupt injection

[Avi Kivity]

  On 08/18/2010 06:32 PM, Mohammed Gamal wrote:
>
>> I did, the disk is not recognized, so something's still broken.
> So far I get hard disk images recognized, but cdrom images aren't.

Try DEBUG_BIOS in pc.c and see why.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.