Re: [PATCH] target/i386: svm: fix sign extension of exit code

[Richard Henderson <richard.henderson@linaro.org>]

On 11/17/25 12:32, Paolo Bonzini wrote:
> On 11/17/25 10:42, Richard Henderson wrote:
>> On 11/15/25 01:26, Paolo Bonzini wrote:
>>> -void cpu_vmexit(CPUX86State *env, uint32_t exit_code, uint64_t exit_info_1,
>>> +void cpu_vmexit(CPUX86State *env, uint64_t exit_code, uint64_t exit_info_1,
>>>                   uintptr_t retaddr)
>>>   {
>>>       CPUState *cs = env_cpu(env);
>>> @@ -732,7 +732,7 @@ void cpu_vmexit(CPUX86State *env, uint32_t exit_code, uint64_t 
>>> exit_info_1,
>>>       qemu_log_mask(CPU_LOG_TB_IN_ASM, "vmexit(%08x, %016" PRIx64 ", %016"
>>>                     PRIx64 ", " TARGET_FMT_lx ")!\n",
>>> -                  exit_code, exit_info_1,
>>> +                  (uint32_t)exit_code, exit_info_1,
>>
>> Why cast instead of printing all 64 bits?
> 
> Because in practice exit_code is either a very small negative value (-1...-4) or a 
> positive value.  For QEMU in addition the positive value will also be small (less than 16 
> bits); values between 0x8000_0000 and 0xffff_ffff could happen in principle but are for 
> use by software and by the processor[1].  So the high 32 bits are basically unused, and 
> the cast removes eight zeroes or f's from the log.

Then maybe you really want the signed int64_t?


r~