Re: [Qemu-devel] [RFC PATCH 1/3] target-i386: Add 486sx, old486, and old486sx CPU models

[Andreas Färber]

Am 28.02.2013 03:12, schrieb H. Peter Anvin:
> From: "H. Peter Anvin" <address@hidden>
> 
> Add models for 486SX, and pre-CPUID versions of the 486 (DX & SX).
> Change the model number for the standard 486DX to a model which
> actually had CPUID.
> 
> Note: these models are fairly vestigial, for example most of the FPU
> operations still work; only F*ST[CS]W have been modified to appear as
> through there is no FPU.
> 
> This also changes the classic 486 model number to 8 (DX4) which
> matches the feature set presented.
> 
> Signed-off-by: H. Peter Anvin <address@hidden>
> ---
>  target-i386/cpu.c         | 39 ++++++++++++++++++++++++++---
>  target-i386/fpu_helper.c  | 12 +++++++--
>  target-i386/misc_helper.c | 15 ++++++++----
>  target-i386/translate.c   | 62 
> +++++++++++++++--------------------------------
>  4 files changed, 75 insertions(+), 53 deletions(-)
> 
> diff --git a/target-i386/cpu.c b/target-i386/cpu.c
> index aab35c7..a5aad19 100644
> --- a/target-i386/cpu.c
> +++ b/target-i386/cpu.c
> @@ -365,8 +365,11 @@ typedef struct x86_def_t {
>      uint32_t cpuid_7_0_ebx_features;
>  } x86_def_t;
>  
> -#define I486_FEATURES (CPUID_FP87 | CPUID_VME | CPUID_PSE)
> -#define PENTIUM_FEATURES (I486_FEATURES | CPUID_DE | CPUID_TSC | \
> +#define OLD_I486SX_FEATURES 0
> +#define OLD_I486_FEATURES CPUID_FP87
> +#define I486SX_FEATURES CPUID_VME /* SX2+ */
> +#define I486_FEATURES (CPUID_FP87 | CPUID_VME) /* DX4 and some DX2 */
> +#define PENTIUM_FEATURES (I486_FEATURES | CPUID_PSE | CPUID_DE | CPUID_TSC | 
> \
>            CPUID_MSR | CPUID_MCE | CPUID_CX8 | CPUID_MMX | CPUID_APIC)
>  #define PENTIUM2_FEATURES (PENTIUM_FEATURES | CPUID_PAE | CPUID_SEP | \
>            CPUID_MTRR | CPUID_PGE | CPUID_MCA | CPUID_CMOV | CPUID_PAT | \
> @@ -535,16 +538,46 @@ static x86_def_t builtin_x86_defs[] = {
>          .model_id = "Genuine Intel(R) CPU           T2600  @ 2.16GHz",
>      },
>      {
> +        .name = "old486",
> +        .level = 0,
> +        .vendor = CPUID_VENDOR_INTEL,
> +        .family = 4,
> +        .model = 1,
> +        .stepping = 0,
> +        .features = OLD_I486_FEATURES,
> +        .xlevel = 0,
> +    },
> +    {
> +        .name = "old486sx",
> +        .level = 0,
> +        .vendor = CPUID_VENDOR_INTEL,
> +        .family = 4,
> +        .model = 2,
> +        .stepping = 0,
> +        .features = OLD_I486SX_FEATURES,
> +        .xlevel = 0,
> +    },
> +    {
>          .name = "486",
>          .level = 1,
>          .vendor = CPUID_VENDOR_INTEL,
>          .family = 4,
> -        .model = 0,
> +        .model = 8,

Such changes have been rejected in the past (e.g., n270 Atom).
I personally wouldn't object to 486 changes, but I guess it should
rather be handled via Igor's CPU static properties that I have in my
review queue: The .model value would be set to 8 but the PC machine
would be changed alongside to set model = 0 for pc-1.4 and earlier.

>          .stepping = 0,
>          .features = I486_FEATURES,
>          .xlevel = 0,
>      },
>      {
> +        .name = "486sx",
> +        .level = 1,
> +        .vendor = CPUID_VENDOR_INTEL,
> +        .family = 4,
> +        .model = 5,
> +        .stepping = 0,
> +        .features = I486SX_FEATURES,
> +        .xlevel = 0,
> +    },
> +    {
>          .name = "pentium",
>          .level = 1,
>          .vendor = CPUID_VENDOR_INTEL,
[...]
> diff --git a/target-i386/translate.c b/target-i386/translate.c
> index 112c310..6d8abff 100644
> --- a/target-i386/translate.c
> +++ b/target-i386/translate.c
[...]
> @@ -7926,6 +7900,8 @@ static inline void 
> gen_intermediate_code_internal(CPUX86State *env,
>      if (flags & HF_SOFTMMU_MASK) {
>          dc->mem_index = (cpu_mmu_index(env) + 1) << 2;
>      }
> +    dc->cpuid_family = (env->cpuid_version >> 8) & 0x0f;
> +    dc->cpuid_level = env->cpuid_level;
>      dc->cpuid_features = env->cpuid_features;
>      dc->cpuid_ext_features = env->cpuid_ext_features;
>      dc->cpuid_ext2_features = env->cpuid_ext2_features;

Would be better to reuse the "family" QOM property that also reads the
upper two nibbles, to avoid surprises.

Regards,
Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg