[PATCH v1 1/2] s390/kvm: split kvm_s390_real_to_abs

[PATCH v1 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>
---
 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 v1 1/2] s390/kvm: split kvm_s390_real_to_abs

[Thomas Huth]

On 19/03/2021 20.33, 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>
> ---
>   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)

<bikeshedding>
Just a matter of taste, but maybe this could be named differently? 
kvm_s390_real2abs_prefix() ? kvm_s390_prefix_real_to_abs()?
</bikeshedding>

Anyway:
Reviewed-by: Thomas Huth <thuth@redhat.com>

Re: [PATCH v1 1/2] s390/kvm: split kvm_s390_real_to_abs

[David Hildenbrand]

On 20.03.21 05:57, Thomas Huth wrote:
> On 19/03/2021 20.33, 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>
>> ---
>>    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)
> 
> <bikeshedding>
> Just a matter of taste, but maybe this could be named differently?
> kvm_s390_real2abs_prefix() ? kvm_s390_prefix_real_to_abs()?
> </bikeshedding>

+1, I also dislike these "_.*" style functions here.

Reviewed-by: David Hildenbrand <david@redhat.com>

> 
> Anyway:
> Reviewed-by: Thomas Huth <thuth@redhat.com>
> 


-- 
Thanks,

David / dhildenb

Re: [PATCH v1 1/2] s390/kvm: split kvm_s390_real_to_abs

[Heiko Carstens]

On Mon, Mar 22, 2021 at 10:53:46AM +0100, David Hildenbrand wrote:
> > > 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)
> > 
> > <bikeshedding>
> > Just a matter of taste, but maybe this could be named differently?
> > kvm_s390_real2abs_prefix() ? kvm_s390_prefix_real_to_abs()?
> > </bikeshedding>
> 
> +1, I also dislike these "_.*" style functions here.

Yes, let's bikeshed then :)

Could you then please try to rename page_to* and everything that looks
similar to page2* please? I'm wondering what the response will be..

Re: [PATCH v1 1/2] s390/kvm: split kvm_s390_real_to_abs

[David Hildenbrand]

On 22.03.21 12:12, Heiko Carstens wrote:
> On Mon, Mar 22, 2021 at 10:53:46AM +0100, David Hildenbrand wrote:
>>>> 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)
>>>
>>> <bikeshedding>
>>> Just a matter of taste, but maybe this could be named differently?
>>> kvm_s390_real2abs_prefix() ? kvm_s390_prefix_real_to_abs()?
>>> </bikeshedding>
>>
>> +1, I also dislike these "_.*" style functions here.
> 
> Yes, let's bikeshed then :)
> 
> Could you then please try to rename page_to* and everything that looks
> similar to page2* please? I'm wondering what the response will be..

Oh, we're bikeshedding about anything now? Cool.

-- 
Thanks,

David / dhildenb

Re: [PATCH v1 1/2] s390/kvm: split kvm_s390_real_to_abs

[David Hildenbrand]

On 22.03.21 12:16, David Hildenbrand wrote:
> On 22.03.21 12:12, Heiko Carstens wrote:
>> On Mon, Mar 22, 2021 at 10:53:46AM +0100, David Hildenbrand wrote:
>>>>> 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)
>>>>
>>>> <bikeshedding>
>>>> Just a matter of taste, but maybe this could be named differently?
>>>> kvm_s390_real2abs_prefix() ? kvm_s390_prefix_real_to_abs()?
>>>> </bikeshedding>
>>>
>>> +1, I also dislike these "_.*" style functions here.
>>
>> Yes, let's bikeshed then :)
>>
>> Could you then please try to rename page_to* and everything that looks
>> similar to page2* please? I'm wondering what the response will be..
> 
> Oh, we're bikeshedding about anything now? Cool.

(I agree that real2abs is not such a good idea ;) )

-- 
Thanks,

David / dhildenb

Re: [PATCH v1 1/2] s390/kvm: split kvm_s390_real_to_abs

[Christian Borntraeger]

On 22.03.21 12:12, Heiko Carstens wrote:
> On Mon, Mar 22, 2021 at 10:53:46AM +0100, David Hildenbrand wrote:
>>>> 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)
>>>
>>> <bikeshedding>
>>> Just a matter of taste, but maybe this could be named differently?
>>> kvm_s390_real2abs_prefix() ? kvm_s390_prefix_real_to_abs()?
>>> </bikeshedding>
>>
>> +1, I also dislike these "_.*" style functions here.
> 
> Yes, let's bikeshed then :)
> 
> Could you then please try to rename page_to* and everything that looks
> similar to page2* please? I'm wondering what the response will be..

Given that this is stable material (due to patch 2), can we try to minimize
the bikeshedding to everything that his touched by this patch?


Claudio, can you respin the series addressing the comments?
I will then either add this to next or fold that into the existing next patches.
Not sure yet.

[PATCH v1 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>
---
 arch/s390/kvm/vsie.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
index 48aab6290a77..92864f9b3d84 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, r1, r2, src, dest, 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 = READ_ONCE(vsie_page->scb_o->mso) & ~(1UL << 20);
+	prefix = scb_s->prefix << GUEST_PREFIX_SHIFT;
 
-	dest = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 16) & mask;
-	src = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 20) & mask;
+	r1 = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 16);
+	r2 = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 20);
+	dest = _kvm_s390_real_to_abs(prefix, r1 & mask) + mso;
+	src = _kvm_s390_real_to_abs(prefix, r2 & mask) + 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 v1 2/2] s390/kvm: VSIE: fix MVPG handling for prefixing and MSO

[David Hildenbrand]

On 19.03.21 20:33, 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>
> ---
>   arch/s390/kvm/vsie.c | 10 +++++++---
>   1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
> index 48aab6290a77..92864f9b3d84 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, r1, r2, src, dest, 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 = READ_ONCE(vsie_page->scb_o->mso) & ~(1UL << 20);

I think we should use the one stored in the shadow page - this is what 
the SIE saw and will see when retrying.

> +	prefix = scb_s->prefix << GUEST_PREFIX_SHIFT;
>   
> -	dest = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 16) & mask;
> -	src = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 20) & mask;
> +	r1 = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 16);
> +	r2 = vsie_get_register(vcpu, vsie_page, scb_s->ipb >> 20);

Just reuse dest and src here?

> +	dest = _kvm_s390_real_to_abs(prefix, r1 & mask) + mso;
> +	src = _kvm_s390_real_to_abs(prefix, r2 & mask) + 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);
> 

Apart from that, looks good to me.

-- 
Thanks,

David / dhildenb