Re: [Qemu-devel] [RFC v1 3/3] target-microblaze: Convert use-fpu to a CPU property

[Peter Crosthwaite]

On Thu, May 14, 2015 at 10:48 PM, Alistair Francis
<address@hidden> wrote:
> On Fri, May 15, 2015 at 3:22 PM, Peter Crosthwaite
> <address@hidden> wrote:
>> On Wed, May 13, 2015 at 11:08 PM, Alistair Francis
>> <address@hidden> wrote:
>>> Originally the use-fpu PVR bits were manually set for each machine. This
>>> is a hassle and difficult to read, instead set them based on the CPU
>>> properties.
>>>
>>> Signed-off-by: Alistair Francis <address@hidden>
>>> ---
>>>
>>>  hw/microblaze/petalogix_ml605_mmu.c |    4 ++--
>>>  target-microblaze/cpu-qom.h         |    1 +
>>>  target-microblaze/cpu.c             |   13 +++++++++++--
>>>  target-microblaze/translate.c       |    6 +++---
>>>  4 files changed, 17 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/hw/microblaze/petalogix_ml605_mmu.c 
>>> b/hw/microblaze/petalogix_ml605_mmu.c
>>> index 48c264b..f4e1cc5 100644
>>> --- a/hw/microblaze/petalogix_ml605_mmu.c
>>> +++ b/hw/microblaze/petalogix_ml605_mmu.c
>>> @@ -71,9 +71,8 @@ static void machine_cpu_reset(MicroBlazeCPU *cpu)
>>>      env->pvr.regs[10] = 0x0e000000; /* virtex 6 */
>>>      /* setup pvr to match kernel setting */
>>>      env->pvr.regs[5] |= PVR5_DCACHE_WRITEBACK_MASK;
>>> -    env->pvr.regs[0] |= PVR0_USE_FPU_MASK | PVR0_ENDI;
>>> +    env->pvr.regs[0] |= PVR0_ENDI;
>>>      env->pvr.regs[0] = (env->pvr.regs[0] & ~PVR0_VERSION_MASK) | (0x14 << 
>>> 8);
>>> -    env->pvr.regs[2] ^= PVR2_USE_FPU2_MASK;
>>
>> So I think this here will clear PVR2_FPU2 ...
>>
>>>      env->pvr.regs[4] = 0xc56b8000;
>>>      env->pvr.regs[5] = 0xc56be000;
>>>  }
>>> @@ -95,6 +94,7 @@ petalogix_ml605_init(MachineState *machine)
>>>
>>>      /* init CPUs */
>>>      cpu = MICROBLAZE_CPU(object_new(TYPE_MICROBLAZE_CPU));
>>> +    object_property_set_int(OBJECT(cpu), 2, "xlnx.use-fpu", &error_abort);
>>
>> But here you are setting use-fpu to 2. Is this a functional change for
>> ML605 board?
>>
>>>      object_property_set_bool(OBJECT(cpu), true, "realized", &error_abort);
>>>
>>>      /* Attach emulated BRAM through the LMB.  */
>>> diff --git a/target-microblaze/cpu-qom.h b/target-microblaze/cpu-qom.h
>>> index 7bc5b81..ce3a574 100644
>>> --- a/target-microblaze/cpu-qom.h
>>> +++ b/target-microblaze/cpu-qom.h
>>> @@ -62,6 +62,7 @@ typedef struct MicroBlazeCPU {
>>>      /* Microblaze Configuration Settings */
>>>      struct {
>>>          bool stackproc;
>>> +        uint8_t usefpu;
>>>      } cfg;
>>>
>>>      CPUMBState env;
>>> diff --git a/target-microblaze/cpu.c b/target-microblaze/cpu.c
>>> index c08da19..c3a9a5d 100644
>>> --- a/target-microblaze/cpu.c
>>> +++ b/target-microblaze/cpu.c
>>> @@ -89,10 +89,18 @@ static void mb_cpu_reset(CPUState *s)
>>>                          | PVR2_USE_DIV_MASK \
>>>                          | PVR2_USE_HW_MUL_MASK \
>>>                          | PVR2_USE_MUL64_MASK \
>>> -                        | PVR2_USE_FPU_MASK \
>>> -                        | PVR2_USE_FPU2_MASK \
>>
>> If I have this straight, this suggests the property default value
>> should be 2, and the ml605 board setting should be 1.
>
> You are correct, I will fix this.
>
>>
>>>                          | PVR2_FPU_EXC_MASK \
>>>                          | 0;
>>> +
>>> +    if (cpu->cfg.usefpu) {
>>> +        env->pvr.regs[0] |= PVR0_USE_FPU_MASK;
>>> +        env->pvr.regs[2] |= PVR2_USE_FPU_MASK;
>>> +
>>> +        if (cpu->cfg.usefpu > 1) {
>>> +            env->pvr.regs[2] |= PVR2_USE_FPU2_MASK;
>>> +        }
>>> +    }
>>
>> This should be handled at realize time. See pl330_realize for example
>> of realize-time PVR ("cfg" in that case) population.
>
> But the env state gets wiped out at reset, so the information will be lost.
>

Ahh, so the solution there is to do what ARM does and have a section
at the back of the env which survives reset. Check the memset in
arm_cpu_reset.

Regards,
Peter

> I can tidy up the if logic though.
>
>>
>> To avoid churn, it might be easiest to convert the PVRs in chunks one
>> by one (e.g. one patch for PVR0, then PVR2 etc etc. The existing |ing
>> can be moved into realize for settings not worth propertyifying.
>>
>>> +
>>>      env->pvr.regs[10] = 0x0c000000; /* Default to spartan 3a dsp family.  
>>> */
>>>      env->pvr.regs[11] = PVR11_USE_MMU | (16 << 17);
>>>
>>> @@ -154,6 +162,7 @@ static Property mb_properties[] = {
>>>      DEFINE_PROP_UINT32("xlnx.base-vectors", MicroBlazeCPU, base_vectors, 
>>> 0),
>>>      DEFINE_PROP_BOOL("xlnx.use-stack-protection", MicroBlazeCPU, 
>>> cfg.stackproc,
>>>                       true),
>>> +    DEFINE_PROP_UINT8("xlnx.use-fpu", MicroBlazeCPU, cfg.usefpu, 1),
>>
>> No need for xlnx prefix (FDT generic should trim the prefix).
>
> I agree, but the base-vectors already had it. If the xlnx is unneeded
> I'll remove
> it from the base-vector as well.
>
> Thanks,
>
> Alistair
>
>>
>> Regards,
>> Peter
>>
>>>      DEFINE_PROP_END_OF_LIST(),
>>>  };
>>>
>>> diff --git a/target-microblaze/translate.c b/target-microblaze/translate.c
>>> index 4068946..36cfc5c 100644
>>> --- a/target-microblaze/translate.c
>>> +++ b/target-microblaze/translate.c
>>> @@ -1415,11 +1415,11 @@ static int dec_check_fpuv2(DisasContext *dc)
>>>
>>>      r = dc->cpu->env.pvr.regs[2] & PVR2_USE_FPU2_MASK;
>>>
>>> -    if (!r && (dc->tb_flags & MSR_EE_FLAG)) {
>>> +    if ((dc->cpu->cfg.usefpu != 2) && (dc->tb_flags & MSR_EE_FLAG)) {
>>>          tcg_gen_movi_tl(cpu_SR[SR_ESR], ESR_EC_FPU);
>>>          t_gen_raise_exception(dc, EXCP_HW_EXCP);
>>>      }
>>> -    return r;
>>> +    return (dc->cpu->cfg.usefpu == 2) ? 0 : PVR2_USE_FPU2_MASK;
>>>  }
>>>
>>>  static void dec_fpu(DisasContext *dc)
>>> @@ -1428,7 +1428,7 @@ static void dec_fpu(DisasContext *dc)
>>>
>>>      if ((dc->tb_flags & MSR_EE_FLAG)
>>>            && (dc->cpu->env.pvr.regs[2] & PVR2_ILL_OPCODE_EXC_MASK)
>>> -          && !((dc->cpu->env.pvr.regs[2] & PVR2_USE_FPU_MASK))) {
>>> +          && (dc->cpu->cfg.usefpu != 1)) {
>>>          tcg_gen_movi_tl(cpu_SR[SR_ESR], ESR_EC_ILLEGAL_OP);
>>>          t_gen_raise_exception(dc, EXCP_HW_EXCP);
>>>          return;
>>> --
>>> 1.7.1
>>>
>>>
>>
>