Re: [PATCH v1 1/5] s390x/mmu: Add EDAT2 translation support

From: Thomas Huth <thuth@redhat.com>
To: David Hildenbrand <david@redhat.com>, qemu-devel@nongnu.org
Cc: Janosch Frank <frankja@linux.ibm.com>,
	Cornelia Huck <cohuck@redhat.com>,
	Halil Pasic <pasic@linux.ibm.com>,
	Christian Borntraeger <borntraeger@de.ibm.com>,
	qemu-s390x@nongnu.org, Richard Henderson <rth@twiddle.net>
Subject: Re: [PATCH v1 1/5] s390x/mmu: Add EDAT2 translation support
Date: Tue, 1 Oct 2019 10:55:04 +0200	[thread overview]
Message-ID: <4fa882e3-04d5-5c73-2266-31b822f1ae3c@redhat.com> (raw)
In-Reply-To: <fc003b6e-a38c-ced7-0c84-a64ebc26e91b@redhat.com>

On 01/10/2019 10.51, David Hildenbrand wrote:
> On 01.10.19 10:41, Thomas Huth wrote:
>> On 26/09/2019 12.18, David Hildenbrand wrote:
>>> On 26.09.19 12:16, David Hildenbrand wrote:
>>>> This only adds basic support to the DAT translation, but no EDAT2 support
>>>> for TCG. E.g., the gdbstub under kvm uses this function, too, to
>>>> translate virtual addresses.
>>>>
>>>> Reviewed-by: Thomas Huth <thuth@redhat.com>
>>>> Signed-off-by: David Hildenbrand <david@redhat.com>
>>>> ---
>>>>  target/s390x/mmu_helper.c | 9 +++++++++
>>>>  1 file changed, 9 insertions(+)
>>>>
>>>> diff --git a/target/s390x/mmu_helper.c b/target/s390x/mmu_helper.c
>>>> index 6b34c4c7b4..54f54137ec 100644
>>>> --- a/target/s390x/mmu_helper.c
>>>> +++ b/target/s390x/mmu_helper.c
>>>> @@ -120,6 +120,7 @@ static int mmu_translate_asce(CPUS390XState *env, target_ulong vaddr,
>>>>  {
>>>>      const bool edat1 = (env->cregs[0] & CR0_EDAT) &&
>>>>                         s390_has_feat(S390_FEAT_EDAT);
>>>> +    const bool edat2 = edat1 && s390_has_feat(S390_FEAT_EDAT_2);
>>>>      const int asce_tl = asce & ASCE_TABLE_LENGTH;
>>>>      const int asce_p = asce & ASCE_PRIVATE_SPACE;
>>>>      hwaddr gaddr = asce & ASCE_ORIGIN;
>>>> @@ -219,9 +220,17 @@ static int mmu_translate_asce(CPUS390XState *env, target_ulong vaddr,
>>>>          if ((entry & REGION_ENTRY_TT) != REGION_ENTRY_TT_REGION3) {
>>>>              return PGM_TRANS_SPEC;
>>>>          }
>>>> +        if (edat2 && (entry & REGION3_ENTRY_CR) && asce_p) {
>>>> +            return PGM_TRANS_SPEC;
>>>> +        }
>>>>          if (edat1 && (entry & REGION_ENTRY_P)) {
>>>>              *flags &= ~PAGE_WRITE;
>>>>          }
>>>> +        if (edat2 && (entry & REGION3_ENTRY_FC)) {
>>>> +            *raddr = (entry & REGION3_ENTRY_RFAA) |
>>>> +                     (vaddr & REGION3_ENTRY_RFAA);
>>>
>>> Messed up
>>>
>>> (vaddr & ~REGION3_ENTRY_RFAA)
>>>
>>> it is.
>>
>> With that fix:
>>
>> Reviewed-by: Thomas Huth <thuth@redhat.com>
>>
> 
> BTW, this change explains the different order of checks you mentioned. I now have here:
> 
> diff --git a/target/s390x/mmu_helper.c b/target/s390x/mmu_helper.c
> index dc33c63b1d..dcbffb682f 100644
> --- a/target/s390x/mmu_helper.c
> +++ b/target/s390x/mmu_helper.c
> @@ -120,6 +120,7 @@ static int mmu_translate_asce(CPUS390XState *env, target_ulong vaddr,
>  {
>      const bool edat1 = (env->cregs[0] & CR0_EDAT) &&
>                         s390_has_feat(S390_FEAT_EDAT);
> +    const bool edat2 = edat1 && s390_has_feat(S390_FEAT_EDAT_2);
>      const int asce_tl = asce & ASCE_TABLE_LENGTH;
>      const int asce_p = asce & ASCE_PRIVATE_SPACE;
>      hwaddr gaddr = asce & ASCE_ORIGIN;
> @@ -217,6 +218,17 @@ static int mmu_translate_asce(CPUS390XState *env, target_ulong vaddr,
>          if ((entry & REGION_ENTRY_TT) != REGION_ENTRY_TT_REGION3) {
>              return PGM_TRANS_SPEC;
>          }
> +        if (edat2 && (entry & REGION3_ENTRY_CR) && asce_p) {
> +            return PGM_TRANS_SPEC;
> +        }
> +        if (edat2 && (entry & REGION3_ENTRY_FC)) {
> +            if (entry & REGION_ENTRY_P) {
> +                *flags &= ~PAGE_WRITE;
> +            }
> +            *raddr = (entry & REGION3_ENTRY_RFAA) |
> +                     (vaddr & ~REGION3_ENTRY_RFAA);
> +            return 0;
> +        }
>          if (VADDR_SEGMENT_TL(vaddr) < (entry & REGION_ENTRY_TF) >> 6 ||
>              VADDR_SEGMENT_TL(vaddr) > (entry & REGION_ENTRY_TL)) {
>              return PGM_SEGMENT_TRANS;

Ah, ok, and the *flags have to be set first, of course. So better keep
it the original way round in your other patch.

 Thomas