To be used after all targets have populated the hook.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/hw/core/cpu.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 2c284d6397..4385ce54c9 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -103,6 +103,8 @@ struct SysemuCPUOps;
  * @parse_features: Callback to parse command line arguments.
  * @reset_dump_flags: #CPUDumpFlags to use for reset logging.
  * @has_work: Callback for checking if there is work to do.
+ * @mmu_index: Callback for choosing softmmu mmu index;
+ *       may be used internally by memory_rw_debug without TCG.
  * @memory_rw_debug: Callback for GDB memory access.
  * @dump_state: Callback for dumping state.
  * @query_cpu_fast:
@@ -150,6 +152,7 @@ struct CPUClass {
     void (*parse_features)(const char *typename, char *str, Error **errp);
 
     bool (*has_work)(CPUState *cpu);
+    int (*mmu_index)(CPUState *cpu, bool ifetch);
     int (*memory_rw_debug)(CPUState *cpu, vaddr addr,
                            uint8_t *buf, int len, bool is_write);
     void (*dump_state)(CPUState *cpu, FILE *, int flags);
-- 
2.34.1

Hi Richard,

On 30/1/24 00:30, Richard Henderson wrote:
> To be used after all targets have populated the hook.
> 
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   include/hw/core/cpu.h | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
> index 2c284d6397..4385ce54c9 100644
> --- a/include/hw/core/cpu.h
> +++ b/include/hw/core/cpu.h
> @@ -103,6 +103,8 @@ struct SysemuCPUOps;
>    * @parse_features: Callback to parse command line arguments.
>    * @reset_dump_flags: #CPUDumpFlags to use for reset logging.
>    * @has_work: Callback for checking if there is work to do.
> + * @mmu_index: Callback for choosing softmmu mmu index;
> + *       may be used internally by memory_rw_debug without TCG.
>    * @memory_rw_debug: Callback for GDB memory access.
>    * @dump_state: Callback for dumping state.
>    * @query_cpu_fast:
> @@ -150,6 +152,7 @@ struct CPUClass {
>       void (*parse_features)(const char *typename, char *str, Error **errp);
>   
>       bool (*has_work)(CPUState *cpu);
> +    int (*mmu_index)(CPUState *cpu, bool ifetch);

Can the index ever be negative?

Would it be useful to also have a mmu_index_max() so we could
check mmu_index() is in range in the final cpu_mmu_index()?

>       int (*memory_rw_debug)(CPUState *cpu, vaddr addr,
>                              uint8_t *buf, int len, bool is_write);
>       void (*dump_state)(CPUState *cpu, FILE *, int flags);

On 1/30/24 17:46, Philippe Mathieu-Daudé wrote:
> Hi Richard,
> 
> On 30/1/24 00:30, Richard Henderson wrote:
>> To be used after all targets have populated the hook.
>>
>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>> ---
>>   include/hw/core/cpu.h | 3 +++
>>   1 file changed, 3 insertions(+)
>>
>> diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
>> index 2c284d6397..4385ce54c9 100644
>> --- a/include/hw/core/cpu.h
>> +++ b/include/hw/core/cpu.h
>> @@ -103,6 +103,8 @@ struct SysemuCPUOps;
>>    * @parse_features: Callback to parse command line arguments.
>>    * @reset_dump_flags: #CPUDumpFlags to use for reset logging.
>>    * @has_work: Callback for checking if there is work to do.
>> + * @mmu_index: Callback for choosing softmmu mmu index;
>> + *       may be used internally by memory_rw_debug without TCG.
>>    * @memory_rw_debug: Callback for GDB memory access.
>>    * @dump_state: Callback for dumping state.
>>    * @query_cpu_fast:
>> @@ -150,6 +152,7 @@ struct CPUClass {
>>       void (*parse_features)(const char *typename, char *str, Error **errp);
>>       bool (*has_work)(CPUState *cpu);
>> +    int (*mmu_index)(CPUState *cpu, bool ifetch);
> 
> Can the index ever be negative?

No.

> Would it be useful to also have a mmu_index_max() so we could
> check mmu_index() is in range in the final cpu_mmu_index()?

We have NB_MMU_MODES, a target-independent constant.


r~

On 30/1/24 08:51, Richard Henderson wrote:
> On 1/30/24 17:46, Philippe Mathieu-Daudé wrote:
>> Hi Richard,
>>
>> On 30/1/24 00:30, Richard Henderson wrote:
>>> To be used after all targets have populated the hook.
>>>
>>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>>> ---
>>>   include/hw/core/cpu.h | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
>>> index 2c284d6397..4385ce54c9 100644
>>> --- a/include/hw/core/cpu.h
>>> +++ b/include/hw/core/cpu.h
>>> @@ -103,6 +103,8 @@ struct SysemuCPUOps;
>>>    * @parse_features: Callback to parse command line arguments.
>>>    * @reset_dump_flags: #CPUDumpFlags to use for reset logging.
>>>    * @has_work: Callback for checking if there is work to do.
>>> + * @mmu_index: Callback for choosing softmmu mmu index;
>>> + *       may be used internally by memory_rw_debug without TCG.
>>>    * @memory_rw_debug: Callback for GDB memory access.
>>>    * @dump_state: Callback for dumping state.
>>>    * @query_cpu_fast:
>>> @@ -150,6 +152,7 @@ struct CPUClass {
>>>       void (*parse_features)(const char *typename, char *str, Error 
>>> **errp);
>>>       bool (*has_work)(CPUState *cpu);
>>> +    int (*mmu_index)(CPUState *cpu, bool ifetch);
>>
>> Can the index ever be negative?
> 
> No.

Then why not return an unsigned type?

>> Would it be useful to also have a mmu_index_max() so we could
>> check mmu_index() is in range in the final cpu_mmu_index()?
> 
> We have NB_MMU_MODES, a target-independent constant.

Ah right.

On 1/30/24 18:20, Philippe Mathieu-Daudé wrote:
>>> Can the index ever be negative?
>>
>> No.
> 
> Then why not return an unsigned type?

The existing function didn't, and I didn't think about it.
Does it really matter?


r~

On 30/1/24 12:02, Richard Henderson wrote:
> On 1/30/24 18:20, Philippe Mathieu-Daudé wrote:
>>>> Can the index ever be negative?
>>>
>>> No.
>>
>> Then why not return an unsigned type?
> 
> The existing function didn't, and I didn't think about it.
> Does it really matter?

No :)