[PATCH v2 0/2] s390/kvm: VSIE: fix prefixing and MSO for MVPG

[PATCH v2 0/2] s390/kvm: VSIE: fix prefixing and MSO for MVPG

[Claudio Imbrenda]

The guest real address needs to pass through prefixing in order to yield
the absolute address.

The absolute address needs to be offset by the MSO in order to get the
host virtual address.

v1->v2
* use the MSO from the shadow SIE page instead of the original one
* reuse src and dest variabled to improve readability

Claudio Imbrenda (2):
  s390/kvm: split kvm_s390_real_to_abs
  s390/kvm: VSIE: fix MVPG handling for prefixing and MSO

 arch/s390/kvm/gaccess.h | 23 +++++++++++++++++------
 arch/s390/kvm/vsie.c    |  6 +++++-
 2 files changed, 22 insertions(+), 7 deletions(-)

-- 
2.26.2

[PATCH v2 1/2] s390/kvm: split kvm_s390_real_to_abs

[Claudio Imbrenda]

A new function _kvm_s390_real_to_abs will apply prefixing to a real address
with a given prefix value.

The old kvm_s390_real_to_abs becomes now a wrapper around the new function.

This is needed to avoid code duplication in vSIE.

Cc: stable@vger.kernel.org
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
---
 arch/s390/kvm/gaccess.h | 23 +++++++++++++++++------
 1 file changed, 17 insertions(+), 6 deletions(-)

diff --git a/arch/s390/kvm/gaccess.h b/arch/s390/kvm/gaccess.h
index daba10f76936..7c72a5e3449f 100644
--- a/arch/s390/kvm/gaccess.h
+++ b/arch/s390/kvm/gaccess.h
@@ -18,17 +18,14 @@
 
 /**
  * kvm_s390_real_to_abs - convert guest real address to guest absolute address
- * @vcpu - guest virtual cpu
+ * @prefix - guest prefix
  * @gra - guest real address
  *
  * Returns the guest absolute address that corresponds to the passed guest real
- * address @gra of a virtual guest cpu by applying its prefix.
+ * address @gra of by applying the given prefix.
  */
-static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu,
-						 unsigned long gra)
+static inline unsigned long _kvm_s390_real_to_abs(u32 prefix, unsigned long gra)
 {
-	unsigned long prefix  = kvm_s390_get_prefix(vcpu);
-
 	if (gra < 2 * PAGE_SIZE)
 		gra += prefix;
 	else if (gra >= prefix && gra < prefix + 2 * PAGE_SIZE)
@@ -36,6 +33,20 @@ static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu,
 	return gra;
 }
 
+/**
+ * kvm_s390_real_to_abs - convert guest real address to guest absolute address
+ * @vcpu - guest virtual cpu
+ * @gra - guest real address
+ *
+ * Returns the guest absolute address that corresponds to the passed guest real
+ * address @gra of a virtual guest cpu by applying its prefix.
+ */
+static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu,
+						 unsigned long gra)
+{
+	return _kvm_s390_real_to_abs(kvm_s390_get_prefix(vcpu), gra);
+}
+
 /**
  * _kvm_s390_logical_to_effective - convert guest logical to effective address
  * @psw: psw of the guest
-- 
2.26.2

Re: [PATCH v2 1/2] s390/kvm: split kvm_s390_real_to_abs

[Christian Borntraeger]

On 22.03.21 15:05, Claudio Imbrenda wrote:
> A new function _kvm_s390_real_to_abs will apply prefixing to a real address
> with a given prefix value.
> 
> The old kvm_s390_real_to_abs becomes now a wrapper around the new function.
> 
> This is needed to avoid code duplication in vSIE.
> 
> Cc: stable@vger.kernel.org
> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
> Reviewed-by: David Hildenbrand <david@redhat.com>
> Reviewed-by: Thomas Huth <thuth@redhat.com>

Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>

> ---
>   arch/s390/kvm/gaccess.h | 23 +++++++++++++++++------
>   1 file changed, 17 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/s390/kvm/gaccess.h b/arch/s390/kvm/gaccess.h
> index daba10f76936..7c72a5e3449f 100644
> --- a/arch/s390/kvm/gaccess.h
> +++ b/arch/s390/kvm/gaccess.h
> @@ -18,17 +18,14 @@
>   
>   /**
>    * kvm_s390_real_to_abs - convert guest real address to guest absolute address
> - * @vcpu - guest virtual cpu
> + * @prefix - guest prefix
>    * @gra - guest real address
>    *
>    * Returns the guest absolute address that corresponds to the passed guest real
> - * address @gra of a virtual guest cpu by applying its prefix.
> + * address @gra of by applying the given prefix.
>    */
> -static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu,
> -						 unsigned long gra)
> +static inline unsigned long _kvm_s390_real_to_abs(u32 prefix, unsigned long gra)
>   {
> -	unsigned long prefix  = kvm_s390_get_prefix(vcpu);
> -
>   	if (gra < 2 * PAGE_SIZE)
>   		gra += prefix;
>   	else if (gra >= prefix && gra < prefix + 2 * PAGE_SIZE)
> @@ -36,6 +33,20 @@ static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu,
>   	return gra;
>   }
>   
> +/**
> + * kvm_s390_real_to_abs - convert guest real address to guest absolute address
> + * @vcpu - guest virtual cpu
> + * @gra - guest real address
> + *
> + * Returns the guest absolute address that corresponds to the passed guest real
> + * address @gra of a virtual guest cpu by applying its prefix.
> + */
> +static inline unsigned long kvm_s390_real_to_abs(struct kvm_vcpu *vcpu,
> +						 unsigned long gra)
> +{
> +	return _kvm_s390_real_to_abs(kvm_s390_get_prefix(vcpu), gra);
> +}
> +
>   /**
>    * _kvm_s390_logical_to_effective - convert guest logical to effective address
>    * @psw: psw of the guest
> 

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

[Claudio Imbrenda]

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);
-- 
2.26.2

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

[David Hildenbrand]

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>

-- 
Thanks,

David / dhildenb

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

[Christian Borntraeger]

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)

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

[David Hildenbrand]

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.

-- 
Thanks,

David / dhildenb

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

[Christian Borntraeger]

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?

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

[David Hildenbrand]

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.

-- 
Thanks,

David / dhildenb

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

[Claudio Imbrenda]

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?

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

[Christian Borntraeger]

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.

Re: [PATCH v2 0/2] s390/kvm: VSIE: fix prefixing and MSO for MVPG

[Christian Borntraeger]

On 22.03.21 15:05, Claudio Imbrenda wrote:
> The guest real address needs to pass through prefixing in order to yield
> the absolute address.
> 
> The absolute address needs to be offset by the MSO in order to get the
> host virtual address.
> 
> v1->v2
> * use the MSO from the shadow SIE page instead of the original one
> * reuse src and dest variabled to improve readability
> 
> Claudio Imbrenda (2):
>    s390/kvm: split kvm_s390_real_to_abs
>    s390/kvm: VSIE: fix MVPG handling for prefixing and MSO
> 
>   arch/s390/kvm/gaccess.h | 23 +++++++++++++++++------
>   arch/s390/kvm/vsie.c    |  6 +++++-
>   2 files changed, 22 insertions(+), 7 deletions(-)
> 

I picked both for internal CI and staging.
RBs or acks can still be send.