diff mbox series

[v2,2/2] s390/kvm: VSIE: fix MVPG handling for prefixing and MSO

Message ID 20210322140559.500716-3-imbrenda@linux.ibm.com (mailing list archive)
State New, archived
Headers show
Series s390/kvm: VSIE: fix prefixing and MSO for MVPG | expand

Commit Message

Claudio Imbrenda March 22, 2021, 2:05 p.m. UTC
Prefixing needs to be applied to the guest real address to translate it
into a guest absolute address.

The value of MSO needs to be added to a guest-absolute address in order to
obtain the host-virtual.

Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when in VSIE")
Cc: stable@vger.kernel.org
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Reported-by: Janosch Frank <frankja@linux.ibm.com>
---
 arch/s390/kvm/vsie.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Comments

David Hildenbrand March 23, 2021, 10:50 a.m. UTC | #1
On 22.03.21 15:05, Claudio Imbrenda wrote:
> Prefixing needs to be applied to the guest real address to translate it
> into a guest absolute address.
> 
> The value of MSO needs to be added to a guest-absolute address in order to
> obtain the host-virtual.
> 
> Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when in VSIE")
> Cc: stable@vger.kernel.org
> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> Reported-by: Janosch Frank <frankja@linux.ibm.com>
> ---
>   arch/s390/kvm/vsie.c | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
> index 48aab6290a77..ac86f11e46dc 100644
> --- a/arch/s390/kvm/vsie.c
> +++ b/arch/s390/kvm/vsie.c
> @@ -1002,7 +1002,7 @@ static u64 vsie_get_register(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page,
>   static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>   {
>   	struct kvm_s390_sie_block *scb_s = &vsie_page->scb_s;
> -	unsigned long pei_dest, pei_src, src, dest, mask;
> +	unsigned long pei_dest, pei_src, dest, src, mask, mso, prefix;
>   	u64 *pei_block = &vsie_page->scb_o->mcic;
>   	int edat, rc_dest, rc_src;
>   	union ctlreg0 cr0;
> @@ -1010,9 +1010,13 @@ static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>   	cr0.val = vcpu->arch.sie_block->gcr[0];
>   	edat = cr0.edat && test_kvm_facility(vcpu->kvm, 8);
>   	mask = _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
> +	mso = scb_s->mso & ~(1UL << 20);
> +	prefix = scb_s->prefix << GUEST_PREFIX_SHIFT;
>   
>   	dest = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 16) & mask;
> +	dest = _kvm_s390_real_to_abs(prefix, dest) + mso;
>   	src = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 20) & mask;
> +	src = _kvm_s390_real_to_abs(prefix, src) + mso;
>   
>   	rc_dest = kvm_s390_shadow_fault(vcpu, vsie_page->gmap, dest, &pei_dest);
>   	rc_src = kvm_s390_shadow_fault(vcpu, vsie_page->gmap, src, &pei_src);
> 

mso is always confusing, but I think this should be correct.

Reviewed-by: David Hildenbrand <david@redhat.com>
Christian Borntraeger March 23, 2021, 3:07 p.m. UTC | #2
On 22.03.21 15:05, Claudio Imbrenda wrote:
> Prefixing needs to be applied to the guest real address to translate it
> into a guest absolute address.
> 
> The value of MSO needs to be added to a guest-absolute address in order to
> obtain the host-virtual.
> 
> Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when in VSIE")
> Cc: stable@vger.kernel.org
> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> Reported-by: Janosch Frank <frankja@linux.ibm.com>
> ---
>   arch/s390/kvm/vsie.c | 6 +++++-
>   1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
> index 48aab6290a77..ac86f11e46dc 100644
> --- a/arch/s390/kvm/vsie.c
> +++ b/arch/s390/kvm/vsie.c
> @@ -1002,7 +1002,7 @@ static u64 vsie_get_register(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page,
>   static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>   {
>   	struct kvm_s390_sie_block *scb_s = &vsie_page->scb_s;
> -	unsigned long pei_dest, pei_src, src, dest, mask;
> +	unsigned long pei_dest, pei_src, dest, src, mask, mso, prefix;
>   	u64 *pei_block = &vsie_page->scb_o->mcic;
>   	int edat, rc_dest, rc_src;
>   	union ctlreg0 cr0;
> @@ -1010,9 +1010,13 @@ static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>   	cr0.val = vcpu->arch.sie_block->gcr[0];
>   	edat = cr0.edat && test_kvm_facility(vcpu->kvm, 8);
>   	mask = _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
> +	mso = scb_s->mso & ~(1UL << 20);
			shouldnt that be ~((1UL << 20 ) -1)
David Hildenbrand March 23, 2021, 3:11 p.m. UTC | #3
On 23.03.21 16:07, Christian Borntraeger wrote:
> 
> 
> On 22.03.21 15:05, Claudio Imbrenda wrote:
>> Prefixing needs to be applied to the guest real address to translate it
>> into a guest absolute address.
>>
>> The value of MSO needs to be added to a guest-absolute address in order to
>> obtain the host-virtual.
>>
>> Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when in VSIE")
>> Cc: stable@vger.kernel.org
>> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
>> Reported-by: Janosch Frank <frankja@linux.ibm.com>
>> ---
>>    arch/s390/kvm/vsie.c | 6 +++++-
>>    1 file changed, 5 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
>> index 48aab6290a77..ac86f11e46dc 100644
>> --- a/arch/s390/kvm/vsie.c
>> +++ b/arch/s390/kvm/vsie.c
>> @@ -1002,7 +1002,7 @@ static u64 vsie_get_register(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page,
>>    static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>    {
>>    	struct kvm_s390_sie_block *scb_s = &vsie_page->scb_s;
>> -	unsigned long pei_dest, pei_src, src, dest, mask;
>> +	unsigned long pei_dest, pei_src, dest, src, mask, mso, prefix;
>>    	u64 *pei_block = &vsie_page->scb_o->mcic;
>>    	int edat, rc_dest, rc_src;
>>    	union ctlreg0 cr0;
>> @@ -1010,9 +1010,13 @@ static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>    	cr0.val = vcpu->arch.sie_block->gcr[0];
>>    	edat = cr0.edat && test_kvm_facility(vcpu->kvm, 8);
>>    	mask = _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
>> +	mso = scb_s->mso & ~(1UL << 20);
> 			shouldnt that be ~((1UL << 20 ) -1)
> 

Looking at shadow_scb(), we can simply take scb_s->mso unmodified.
Christian Borntraeger March 23, 2021, 3:16 p.m. UTC | #4
On 23.03.21 16:11, David Hildenbrand wrote:
> On 23.03.21 16:07, Christian Borntraeger wrote:
>>
>>
>> On 22.03.21 15:05, Claudio Imbrenda wrote:
>>> Prefixing needs to be applied to the guest real address to translate it
>>> into a guest absolute address.
>>>
>>> The value of MSO needs to be added to a guest-absolute address in order to
>>> obtain the host-virtual.
>>>
>>> Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when in VSIE")
>>> Cc: stable@vger.kernel.org
>>> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
>>> Reported-by: Janosch Frank <frankja@linux.ibm.com>
>>> ---
>>>    arch/s390/kvm/vsie.c | 6 +++++-
>>>    1 file changed, 5 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
>>> index 48aab6290a77..ac86f11e46dc 100644
>>> --- a/arch/s390/kvm/vsie.c
>>> +++ b/arch/s390/kvm/vsie.c
>>> @@ -1002,7 +1002,7 @@ static u64 vsie_get_register(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page,
>>>    static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>>    {
>>>        struct kvm_s390_sie_block *scb_s = &vsie_page->scb_s;
>>> -    unsigned long pei_dest, pei_src, src, dest, mask;
>>> +    unsigned long pei_dest, pei_src, dest, src, mask, mso, prefix;
>>>        u64 *pei_block = &vsie_page->scb_o->mcic;
>>>        int edat, rc_dest, rc_src;
>>>        union ctlreg0 cr0;
>>> @@ -1010,9 +1010,13 @@ static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>>        cr0.val = vcpu->arch.sie_block->gcr[0];
>>>        edat = cr0.edat && test_kvm_facility(vcpu->kvm, 8);
>>>        mask = _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
>>> +    mso = scb_s->mso & ~(1UL << 20);
>>             shouldnt that be ~((1UL << 20 ) -1)
>>
> 
> Looking at shadow_scb(), we can simply take scb_s->mso unmodified.

Right, I think I can fix this up (and get rid of the local mso variable)
when queueing. Or shall Claudio send a new version of the patch?
David Hildenbrand March 23, 2021, 3:17 p.m. UTC | #5
On 23.03.21 16:16, Christian Borntraeger wrote:
> 
> 
> On 23.03.21 16:11, David Hildenbrand wrote:
>> On 23.03.21 16:07, Christian Borntraeger wrote:
>>>
>>>
>>> On 22.03.21 15:05, Claudio Imbrenda wrote:
>>>> Prefixing needs to be applied to the guest real address to translate it
>>>> into a guest absolute address.
>>>>
>>>> The value of MSO needs to be added to a guest-absolute address in order to
>>>> obtain the host-virtual.
>>>>
>>>> Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when in VSIE")
>>>> Cc: stable@vger.kernel.org
>>>> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
>>>> Reported-by: Janosch Frank <frankja@linux.ibm.com>
>>>> ---
>>>>     arch/s390/kvm/vsie.c | 6 +++++-
>>>>     1 file changed, 5 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
>>>> index 48aab6290a77..ac86f11e46dc 100644
>>>> --- a/arch/s390/kvm/vsie.c
>>>> +++ b/arch/s390/kvm/vsie.c
>>>> @@ -1002,7 +1002,7 @@ static u64 vsie_get_register(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page,
>>>>     static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>>>     {
>>>>         struct kvm_s390_sie_block *scb_s = &vsie_page->scb_s;
>>>> -    unsigned long pei_dest, pei_src, src, dest, mask;
>>>> +    unsigned long pei_dest, pei_src, dest, src, mask, mso, prefix;
>>>>         u64 *pei_block = &vsie_page->scb_o->mcic;
>>>>         int edat, rc_dest, rc_src;
>>>>         union ctlreg0 cr0;
>>>> @@ -1010,9 +1010,13 @@ static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>>>         cr0.val = vcpu->arch.sie_block->gcr[0];
>>>>         edat = cr0.edat && test_kvm_facility(vcpu->kvm, 8);
>>>>         mask = _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
>>>> +    mso = scb_s->mso & ~(1UL << 20);
>>>              shouldnt that be ~((1UL << 20 ) -1)
>>>
>>
>> Looking at shadow_scb(), we can simply take scb_s->mso unmodified.
> 
> Right, I think I can fix this up (and get rid of the local mso variable)
> when queueing. Or shall Claudio send a new version of the patch?
> 

IMHO, you can just fix it up.
Claudio Imbrenda March 23, 2021, 3:21 p.m. UTC | #6
On Tue, 23 Mar 2021 16:16:18 +0100
Christian Borntraeger <borntraeger@de.ibm.com> wrote:

> On 23.03.21 16:11, David Hildenbrand wrote:
> > On 23.03.21 16:07, Christian Borntraeger wrote:  
> >>
> >>
> >> On 22.03.21 15:05, Claudio Imbrenda wrote:  
> >>> Prefixing needs to be applied to the guest real address to
> >>> translate it into a guest absolute address.
> >>>
> >>> The value of MSO needs to be added to a guest-absolute address in
> >>> order to obtain the host-virtual.
> >>>
> >>> Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when
> >>> in VSIE") Cc: stable@vger.kernel.org
> >>> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> >>> Reported-by: Janosch Frank <frankja@linux.ibm.com>
> >>> ---
> >>>    arch/s390/kvm/vsie.c | 6 +++++-
> >>>    1 file changed, 5 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
> >>> index 48aab6290a77..ac86f11e46dc 100644
> >>> --- a/arch/s390/kvm/vsie.c
> >>> +++ b/arch/s390/kvm/vsie.c
> >>> @@ -1002,7 +1002,7 @@ static u64 vsie_get_register(struct
> >>> kvm_vcpu *vcpu, struct vsie_page *vsie_page, static int
> >>> vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page
> >>> *vsie_page) { struct kvm_s390_sie_block *scb_s =
> >>> &vsie_page->scb_s;
> >>> -    unsigned long pei_dest, pei_src, src, dest, mask;
> >>> +    unsigned long pei_dest, pei_src, dest, src, mask, mso,
> >>> prefix; u64 *pei_block = &vsie_page->scb_o->mcic;
> >>>        int edat, rc_dest, rc_src;
> >>>        union ctlreg0 cr0;
> >>> @@ -1010,9 +1010,13 @@ static int vsie_handle_mvpg(struct
> >>> kvm_vcpu *vcpu, struct vsie_page *vsie_page) cr0.val =
> >>> vcpu->arch.sie_block->gcr[0]; edat = cr0.edat &&
> >>> test_kvm_facility(vcpu->kvm, 8); mask =
> >>> _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
> >>> +    mso = scb_s->mso & ~(1UL << 20);  
> >>             shouldnt that be ~((1UL << 20 ) -1)

oops

> > 
> > Looking at shadow_scb(), we can simply take scb_s->mso unmodified.  
> 
> Right, I think I can fix this up (and get rid of the local mso

I think that's easier/simpler

> variable) when queueing. Or shall Claudio send a new version of the
> patch?
Christian Borntraeger March 23, 2021, 3:34 p.m. UTC | #7
On 23.03.21 16:21, Claudio Imbrenda wrote:
> On Tue, 23 Mar 2021 16:16:18 +0100
> Christian Borntraeger <borntraeger@de.ibm.com> wrote:
> 
>> On 23.03.21 16:11, David Hildenbrand wrote:
>>> On 23.03.21 16:07, Christian Borntraeger wrote:
>>>>
>>>>
>>>> On 22.03.21 15:05, Claudio Imbrenda wrote:
>>>>> Prefixing needs to be applied to the guest real address to
>>>>> translate it into a guest absolute address.
>>>>>
>>>>> The value of MSO needs to be added to a guest-absolute address in
>>>>> order to obtain the host-virtual.
>>>>>
>>>>> Fixes: 223ea46de9e79 ("s390/kvm: VSIE: correctly handle MVPG when
>>>>> in VSIE") Cc: stable@vger.kernel.org
>>>>> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
>>>>> Reported-by: Janosch Frank <frankja@linux.ibm.com>
>>>>> ---
>>>>>     arch/s390/kvm/vsie.c | 6 +++++-
>>>>>     1 file changed, 5 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
>>>>> index 48aab6290a77..ac86f11e46dc 100644
>>>>> --- a/arch/s390/kvm/vsie.c
>>>>> +++ b/arch/s390/kvm/vsie.c
>>>>> @@ -1002,7 +1002,7 @@ static u64 vsie_get_register(struct
>>>>> kvm_vcpu *vcpu, struct vsie_page *vsie_page, static int
>>>>> vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page
>>>>> *vsie_page) { struct kvm_s390_sie_block *scb_s =
>>>>> &vsie_page->scb_s;
>>>>> -    unsigned long pei_dest, pei_src, src, dest, mask;
>>>>> +    unsigned long pei_dest, pei_src, dest, src, mask, mso,
>>>>> prefix; u64 *pei_block = &vsie_page->scb_o->mcic;
>>>>>         int edat, rc_dest, rc_src;
>>>>>         union ctlreg0 cr0;
>>>>> @@ -1010,9 +1010,13 @@ static int vsie_handle_mvpg(struct
>>>>> kvm_vcpu *vcpu, struct vsie_page *vsie_page) cr0.val =
>>>>> vcpu->arch.sie_block->gcr[0]; edat = cr0.edat &&
>>>>> test_kvm_facility(vcpu->kvm, 8); mask =
>>>>> _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
>>>>> +    mso = scb_s->mso & ~(1UL << 20);
>>>>              shouldnt that be ~((1UL << 20 ) -1)
> 
> oops
> 
>>>
>>> Looking at shadow_scb(), we can simply take scb_s->mso unmodified.
>>
>> Right, I think I can fix this up (and get rid of the local mso
> 
> I think that's easier/simpler
> 
>> variable) when queueing. Or shall Claudio send a new version of the
>> patch?
> 

queued (with fixup). Please note: can you use "KVM: s390:" instead of "s390/kvm" for the future patches.
diff mbox series

Patch

diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
index 48aab6290a77..ac86f11e46dc 100644
--- a/arch/s390/kvm/vsie.c
+++ b/arch/s390/kvm/vsie.c
@@ -1002,7 +1002,7 @@  static u64 vsie_get_register(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page,
 static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
 {
 	struct kvm_s390_sie_block *scb_s = &vsie_page->scb_s;
-	unsigned long pei_dest, pei_src, src, dest, mask;
+	unsigned long pei_dest, pei_src, dest, src, mask, mso, prefix;
 	u64 *pei_block = &vsie_page->scb_o->mcic;
 	int edat, rc_dest, rc_src;
 	union ctlreg0 cr0;
@@ -1010,9 +1010,13 @@  static int vsie_handle_mvpg(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
 	cr0.val = vcpu->arch.sie_block->gcr[0];
 	edat = cr0.edat && test_kvm_facility(vcpu->kvm, 8);
 	mask = _kvm_s390_logical_to_effective(&scb_s->gpsw, PAGE_MASK);
+	mso = scb_s->mso & ~(1UL << 20);
+	prefix = scb_s->prefix << GUEST_PREFIX_SHIFT;
 
 	dest = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 16) & mask;
+	dest = _kvm_s390_real_to_abs(prefix, dest) + mso;
 	src = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 20) & mask;
+	src = _kvm_s390_real_to_abs(prefix, src) + mso;
 
 	rc_dest = kvm_s390_shadow_fault(vcpu, vsie_page->gmap, dest, &pei_dest);
 	rc_src = kvm_s390_shadow_fault(vcpu, vsie_page->gmap, src, &pei_src);