[Qemu-devel] [PATCH 0/2] s390x: virtio-less ccw target enablement

[Qemu-devel] [PATCH 0/2] s390x: virtio-less ccw target enablement

[Alexander Graf]

Hi,

With these two patches I'm able to boot a kernel with -M s390-ccw and TCG. So
far the actual CCW instruction interpretation is missing, so it won't get any
further than an initrd, but at least it's a start ;).


Alex

Alexander Graf (2):
  s390x: Fix sclp console input
  s390x: Ignore sam31

 target-s390x/insn-data.def | 4 +++-
 target-s390x/interrupt.c   | 2 --
 2 files changed, 3 insertions(+), 3 deletions(-)

-- 
1.7.12.4

[Qemu-devel] [PATCH 1/2] s390x: Fix sclp console input

[Alexander Graf]

When injecting an sclp console interrupt into the guest, we increase
the PC by 4 for some reason. I have no idea why I put that code there,
but it's clearly wrong. Remove the increment.

This patch fixes sclp serial input for the ccw machine.

Signed-off-by: Alexander Graf <agraf@suse.de>
---
 target-s390x/interrupt.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/target-s390x/interrupt.c b/target-s390x/interrupt.c
index 23a9114..1404d0a 100644
--- a/target-s390x/interrupt.c
+++ b/target-s390x/interrupt.c
@@ -22,9 +22,7 @@ void s390_sclp_extint(uint32_t parm)
         kvm_s390_service_interrupt(parm);
     } else {
         S390CPU *dummy_cpu = s390_cpu_addr2state(0);
-        CPUS390XState *env = &dummy_cpu->env;
 
-        env->psw.addr += 4;
         cpu_inject_ext(dummy_cpu, EXT_SERVICE, parm, 0);
     }
 }
-- 
1.7.12.4

[Qemu-devel] [PATCH 2/2] s390x: Ignore sam31

[Alexander Graf]

The SAM31 instruction switches the memory model to 31bit. Linux uses
it on bootup, so we have to implement it. We only really model the 64bit
memory model though, so we can nop it, cross our fingers and hope for
the guest to not rely on 32bit wrap semantics.

Signed-off-by: Alexander Graf <agraf@suse.de>
---
 target-s390x/insn-data.def | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/target-s390x/insn-data.def b/target-s390x/insn-data.def
index b42ebb6..4cc2da4 100644
--- a/target-s390x/insn-data.def
+++ b/target-s390x/insn-data.def
@@ -744,8 +744,10 @@
 /* SERVICE CALL LOGICAL PROCESSOR (PV hypercall) */
     C(0xb220, SERVC,   RRE,   Z,   r1_o, r2_o, 0, 0, servc, 0)
 /* SET ADDRESSING MODE */
+    /* We have to nop SAM31, it is used by early Linux bootup */
+    C(0x010d, SAM31,   E,     Z,   0, 0, 0, 0, 0, 0)
     /* We only do 64-bit, so accept this as a no-op.
-       Let SAM24 and SAM31 signal illegal instruction.  */
+       Let SAM24 signal illegal instruction.  */
     C(0x010e, SAM64,   E,     Z,   0, 0, 0, 0, 0, 0)
 /* SET ADDRESS SPACE CONTROL FAST */
     C(0xb279, SACF,    S,     Z,   0, a2, 0, 0, sacf, 0)
-- 
1.7.12.4

Re: [Qemu-devel] [PATCH 2/2] s390x: Ignore sam31

[Bastian Koppelmann]

Hi Alex,

I'm not really happy with that. A brief look into the instruction manual 
shows, that modal instructions have different outputs depending on the 
addressing mode. On the long run it would be better to implement this. 
That said I'm not an expert on s390x and there is a reason we're only 
doing 64-bit.

Thanks,
Bastian

On 11/04/2014 12:21 PM, Alexander Graf wrote:
> The SAM31 instruction switches the memory model to 31bit. Linux uses
> it on bootup, so we have to implement it. We only really model the 64bit
> memory model though, so we can nop it, cross our fingers and hope for
> the guest to not rely on 32bit wrap semantics.
>
> Signed-off-by: Alexander Graf <agraf@suse.de>
> ---
>   target-s390x/insn-data.def | 4 +++-
>   1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/target-s390x/insn-data.def b/target-s390x/insn-data.def
> index b42ebb6..4cc2da4 100644
> --- a/target-s390x/insn-data.def
> +++ b/target-s390x/insn-data.def
> @@ -744,8 +744,10 @@
>   /* SERVICE CALL LOGICAL PROCESSOR (PV hypercall) */
>       C(0xb220, SERVC,   RRE,   Z,   r1_o, r2_o, 0, 0, servc, 0)
>   /* SET ADDRESSING MODE */
> +    /* We have to nop SAM31, it is used by early Linux bootup */
> +    C(0x010d, SAM31,   E,     Z,   0, 0, 0, 0, 0, 0)
>       /* We only do 64-bit, so accept this as a no-op.
> -       Let SAM24 and SAM31 signal illegal instruction.  */
> +       Let SAM24 signal illegal instruction.  */
>       C(0x010e, SAM64,   E,     Z,   0, 0, 0, 0, 0, 0)
>   /* SET ADDRESS SPACE CONTROL FAST */
>       C(0xb279, SACF,    S,     Z,   0, a2, 0, 0, sacf, 0)

Re: [Qemu-devel] [PATCH 1/2] s390x: Fix sclp console input

[Bastian Koppelmann]

On 11/04/2014 12:21 PM, Alexander Graf wrote:
> When injecting an sclp console interrupt into the guest, we increase
> the PC by 4 for some reason. I have no idea why I put that code there,
> but it's clearly wrong. Remove the increment.
>
> This patch fixes sclp serial input for the ccw machine.
>
> Signed-off-by: Alexander Graf <agraf@suse.de>
> ---
>   target-s390x/interrupt.c | 2 --
>   1 file changed, 2 deletions(-)
>
> diff --git a/target-s390x/interrupt.c b/target-s390x/interrupt.c
> index 23a9114..1404d0a 100644
> --- a/target-s390x/interrupt.c
> +++ b/target-s390x/interrupt.c
> @@ -22,9 +22,7 @@ void s390_sclp_extint(uint32_t parm)
>           kvm_s390_service_interrupt(parm);
>       } else {
>           S390CPU *dummy_cpu = s390_cpu_addr2state(0);
> -        CPUS390XState *env = &dummy_cpu->env;
>   
> -        env->psw.addr += 4;
>           cpu_inject_ext(dummy_cpu, EXT_SERVICE, parm, 0);
>       }
>   }
Reviewed-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>

Re: [Qemu-devel] [PATCH 2/2] s390x: Ignore sam31

[Alexander Graf]

On 04.11.14 19:40, Bastian Koppelmann wrote:
> Hi Alex,
> 
> I'm not really happy with that. A brief look into the instruction manual
> shows, that modal instructions have different outputs depending on the
> addressing mode. On the long run it would be better to implement this.
> That said I'm not an expert on s390x and there is a reason we're only
> doing 64-bit.

IIRC we even do support 31bit mode - we merely would need to set the
respective PSW bit :).

I'll give that a try.


Alex