Re: [Qemu-ppc] [Qemu-devel] [PATCH] target-ppc: Init dcache and icache size for e500 user mode

[Scott Wood]

On 04/18/2012 04:24 PM, Alexander Graf wrote:
> 
> On 18.04.2012, at 18:27, Scott Wood wrote:
> 
>> On 04/18/2012 10:40 AM, Alexander Graf wrote:
>>> On 04/15/2012 06:14 PM, Andreas Färber wrote:
>>>> diff --git a/target-ppc/translate_init.c b/target-ppc/translate_init.c
>>>> index 86a915c..ba4b84d 100644
>>>> --- a/target-ppc/translate_init.c
>>>> +++ b/target-ppc/translate_init.c
>>>> @@ -4462,36 +4462,32 @@ static void init_proc_e500 (CPUPPCState *env,
>>>> int version)
>>>>                   &spr_read_spefscr,&spr_write_spefscr,
>>>>                   0x00000000);
>>>>      /* Memory management */
>>>> -#if !defined(CONFIG_USER_ONLY)
>>>> +#if defined(CONFIG_USER_ONLY)
>>>> +    env->dcache_line_size = 32;
>>>> +    env->icache_line_size = 32;
>>>> +#else /* !defined(CONFIG_USER_ONLY) */
>>>
>>> I don't think I like this one. If you're running linux-user with a
>>> specific CPU (not the generic "I am the common ground" CPU)
>>
>> But how much value is there really in doing that, that justifies a bunch
>> of ifdefs in the per-CPU code?
> 
> As much value as in allowing specific CPUs to be selected in the first place.

Right, I was wondering how much sense it makes to select a particular
CPU for user-only, or whether it might make more sense to just select
particular user-visible features.

The right answer is probably just to rework all this so that it's more
data-driven and we don't need so many ifdefs regardless.

-Scott