Let's also put the 31-bit hack in front of the REAL MMU, otherwise right
now we get errors when loading a PSW where the highest bit is set (e.g.
via s390-netboot.img). The highest bit is not masked away, therefore we
inject addressing exceptions into the guest.

The proper fix will later be to do all address wrapping before accessing
the MMU - so we won't get any "wrong" entries in there (which makes
flushing also easier). But that will require more work (wrapping in
load_psw, wrapping when incrementing the PC, wrapping every memory
access).

This fixes the tests/pxe-test test.

Signed-off-by: David Hildenbrand <david@redhat.com>
---
 target/s390x/excp_helper.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/target/s390x/excp_helper.c b/target/s390x/excp_helper.c
index 411051edc3..dfee221111 100644
--- a/target/s390x/excp_helper.c
+++ b/target/s390x/excp_helper.c
@@ -107,6 +107,10 @@ int s390_cpu_handle_mmu_fault(CPUState *cs, vaddr orig_vaddr, int size,
             return 1;
         }
     } else if (mmu_idx == MMU_REAL_IDX) {
+        /* 31-Bit mode */
+        if (!(env->psw.mask & PSW_MASK_64)) {
+            vaddr &= 0x7fffffff;
+        }
         if (mmu_translate_real(env, vaddr, rw, &raddr, &prot)) {
             return 1;
         }
-- 
2.14.3

On Thu,  1 Mar 2018 13:08:26 +0100
David Hildenbrand <david@redhat.com> wrote:

> Let's also put the 31-bit hack in front of the REAL MMU, otherwise right
> now we get errors when loading a PSW where the highest bit is set (e.g.
> via s390-netboot.img). The highest bit is not masked away, therefore we
> inject addressing exceptions into the guest.
> 
> The proper fix will later be to do all address wrapping before accessing
> the MMU - so we won't get any "wrong" entries in there (which makes
> flushing also easier). But that will require more work (wrapping in
> load_psw, wrapping when incrementing the PC, wrapping every memory
> access).
> 
> This fixes the tests/pxe-test test.
> 
> Signed-off-by: David Hildenbrand <david@redhat.com>
> ---
>  target/s390x/excp_helper.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/target/s390x/excp_helper.c b/target/s390x/excp_helper.c
> index 411051edc3..dfee221111 100644
> --- a/target/s390x/excp_helper.c
> +++ b/target/s390x/excp_helper.c
> @@ -107,6 +107,10 @@ int s390_cpu_handle_mmu_fault(CPUState *cs, vaddr orig_vaddr, int size,
>              return 1;
>          }
>      } else if (mmu_idx == MMU_REAL_IDX) {
> +        /* 31-Bit mode */
> +        if (!(env->psw.mask & PSW_MASK_64)) {
> +            vaddr &= 0x7fffffff;
> +        }
>          if (mmu_translate_real(env, vaddr, rw, &raddr, &prot)) {
>              return 1;
>          }

Thanks, queued to s390-next.

On 01.03.2018 13:08, David Hildenbrand wrote:
> Let's also put the 31-bit hack in front of the REAL MMU, otherwise right
> now we get errors when loading a PSW where the highest bit is set (e.g.
> via s390-netboot.img). The highest bit is not masked away, therefore we
> inject addressing exceptions into the guest.
> 
> The proper fix will later be to do all address wrapping before accessing
> the MMU - so we won't get any "wrong" entries in there (which makes
> flushing also easier). But that will require more work (wrapping in
> load_psw, wrapping when incrementing the PC, wrapping every memory
> access).
> 
> This fixes the tests/pxe-test test.
> 
> Signed-off-by: David Hildenbrand <david@redhat.com>
> ---
>  target/s390x/excp_helper.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/target/s390x/excp_helper.c b/target/s390x/excp_helper.c
> index 411051edc3..dfee221111 100644
> --- a/target/s390x/excp_helper.c
> +++ b/target/s390x/excp_helper.c
> @@ -107,6 +107,10 @@ int s390_cpu_handle_mmu_fault(CPUState *cs, vaddr orig_vaddr, int size,
>              return 1;
>          }
>      } else if (mmu_idx == MMU_REAL_IDX) {
> +        /* 31-Bit mode */
> +        if (!(env->psw.mask & PSW_MASK_64)) {
> +            vaddr &= 0x7fffffff;
> +        }

Since the preceeding if-statement has exactly the same check, I think
you could also merge the two checks by putting that in front of the
if-statement instead?

Apart from that, patch looks good to me.

 Thomas

On 01.03.2018 14:53, Thomas Huth wrote:
> On 01.03.2018 13:08, David Hildenbrand wrote:
>> Let's also put the 31-bit hack in front of the REAL MMU, otherwise right
>> now we get errors when loading a PSW where the highest bit is set (e.g.
>> via s390-netboot.img). The highest bit is not masked away, therefore we
>> inject addressing exceptions into the guest.
>>
>> The proper fix will later be to do all address wrapping before accessing
>> the MMU - so we won't get any "wrong" entries in there (which makes
>> flushing also easier). But that will require more work (wrapping in
>> load_psw, wrapping when incrementing the PC, wrapping every memory
>> access).
>>
>> This fixes the tests/pxe-test test.
>>
>> Signed-off-by: David Hildenbrand <david@redhat.com>
>> ---
>>  target/s390x/excp_helper.c | 4 ++++
>>  1 file changed, 4 insertions(+)
>>
>> diff --git a/target/s390x/excp_helper.c b/target/s390x/excp_helper.c
>> index 411051edc3..dfee221111 100644
>> --- a/target/s390x/excp_helper.c
>> +++ b/target/s390x/excp_helper.c
>> @@ -107,6 +107,10 @@ int s390_cpu_handle_mmu_fault(CPUState *cs, vaddr orig_vaddr, int size,
>>              return 1;
>>          }
>>      } else if (mmu_idx == MMU_REAL_IDX) {
>> +        /* 31-Bit mode */
>> +        if (!(env->psw.mask & PSW_MASK_64)) {
>> +            vaddr &= 0x7fffffff;
>> +        }
> 
> Since the preceeding if-statement has exactly the same check, I think
> you could also merge the two checks by putting that in front of the
> if-statement instead?
> 
> Apart from that, patch looks good to me.
>

Yes, but as it is already submitted in a pull request I guess we'll
leave it that way.

Want to get rid off all that hacky stuff soon (once I have some spare
time ...)

Thanks!

>  Thomas
> 
-- 

Thanks,

David / dhildenb