Re: [Qemu-devel] [PATCH v4 0/4] semihosting at translate time fixes

[Alex Bennée]

Peter Maydell <address@hidden> writes:

> On Fri, 6 Sep 2019 at 21:26, Alex Bennée <address@hidden> wrote:
>>
>> Hi Peter,
>>
>> Hopefully this is the final version of the semihosting at translate
>> time patches. I've applied Richard's IS_USER changes and gated the SVN
>> for !M profile.
>>
>> Alex Bennée (3):
>>   target/arm: handle M-profile semihosting at translate time
>>   target/arm: handle A-profile semihosting at translate time
>>   target/arm: remove run time semihosting checks
>
> Hi. I've just been looking at these, and I noticed that
> they seem to accidentally extend the "no semihosting
> in user mode" check that is currently for softmmu only
> to also cover linux-user mode (where it would amount
> to "never provide semihosting").

I misread Richard's comments - he only actually said to drop the #ifndef
CONFIG_USER while using !IS_USER for M profile. I'll return the #ifndef
CONFIG_USER for A-profile.

It does seem a bit weird that userspace linux-user does do semihosting
whereas EL0 in softmmu doesn't. Is that because we are effectively
short-circuiting what a real ARM kernel would be doing for EL0?

> This is because we used
> to do the check in the helper.c code which is only used
> by softmmu, and not in the linux-user/arm/cpu_loop.c
> equivalent that linux-user uses. But now we do the check
> in translate.c, which is common to both.
>
> There's also some missed cleanup in that the linux-user
> code can also have the "maybe EXCP_BKPT/EXCP_SWI is a semihosting
> call" checks deleted.

I'll have a look at that.

>
>> Emilio G. Cota (1):
>>   atomic_template: fix indentation in GEN_ATOMIC_HELPER
>
> I've taken the atomic_template fix into target-arm.next,
> since it's unrelated.
>
> thanks
> -- PMM


--
Alex Bennée