Re: [RFC v12 55/65] target/arm: cpu-exceptions: new module

[Claudio Fontana]

On 3/28/21 8:40 PM, Richard Henderson wrote:
> On 3/26/21 1:36 PM, Claudio Fontana wrote:
>> extract the exception handling code from cpu-sysemu,
>> and split it into general arm code and an AArch64-specific part.
>>
>> Signed-off-by: Claudio Fontana<cfontana@suse.de>
>> ---
>>   target/arm/cpu-exceptions-aa64.h |  32 +
>>   target/arm/cpu.h                 |   4 -
>>   target/arm/arch_dump.c           |   1 +
>>   target/arm/cpu-exceptions-aa64.c | 553 ++++++++++++++++++
>>   target/arm/cpu-exceptions.c      | 481 +++++++++++++++
>>   target/arm/cpu-sysemu.c          | 975 -------------------------------
>>   target/arm/cpu-user.c            |   1 +
>>   target/arm/cpu64.c               |   1 +
>>   target/arm/kvm/kvm64.c           |   1 +
>>   target/arm/tcg/helper-a64.c      |   1 +
>>   target/arm/tcg/helper.c          |   1 +
>>   target/arm/meson.build           |   5 +
>>   12 files changed, 1077 insertions(+), 979 deletions(-)
>>   create mode 100644 target/arm/cpu-exceptions-aa64.h
>>   create mode 100644 target/arm/cpu-exceptions-aa64.c
>>   create mode 100644 target/arm/cpu-exceptions.c
> 
> This is the second move for all of this code.
> Not a fan of that.  Come up with a plan and move it once.

I tried to reduce the double move, I failed at my first attempt: I tried to 
move the exception code directly in place
but failed due to first the code needing to be extracted out from being 
tcg-only fairly early to enable the build,
but then the code in exceptions.c / exceptions-aa64.c needing the cpu-sve stuff 
and the aa64 split of other patches.


> 
> I think the sve_exception_el move is a mistake, since as you have pointed out 
> it is used far beyond exception handling.

I think that's right, also switch_mode and the _sync_ functions probably do not 
belong there.. I'll revisit this.

> 
> 
> r~
>
 
Thanks,

C