Re: [Qemu-devel] [PATCH 3/5] arm: Support thumb in set_pc routines

[Peter Crosthwaite]

On Mon, Jun 15, 2015 at 12:36 AM, Peter Maydell
<address@hidden> wrote:
> On 15 June 2015 at 04:48, Peter Crosthwaite <address@hidden> wrote:
>> ARM program counters are always at least 16b aligned with the LSB
>> being only used the indicate thumb mode in exchange situations. Mask
>> this bit off in set_pc to ignore the exchange semantic (which must
>> still be managed by the caller).
>>
>> Signed-off-by: Peter Crosthwaite <address@hidden>
>> ---
>> ---
>>  target-arm/cpu.c   | 2 +-
>>  target-arm/cpu64.c | 2 +-
>>  2 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/target-arm/cpu.c b/target-arm/cpu.c
>> index 6181f28..5bb08a6 100644
>> --- a/target-arm/cpu.c
>> +++ b/target-arm/cpu.c
>> @@ -35,7 +35,7 @@ static void arm_cpu_set_pc(CPUState *cs, vaddr value)
>>  {
>>      ARMCPU *cpu = ARM_CPU(cs);
>>
>> -    cpu->env.regs[15] = value;
>> +    cpu->env.regs[15] = value & 0xfffffffe;
>>  }
>
>
> This doesn't look right to me. There are two semantics that
> make sense for setting an ARM PC value:
>
>  (1) interworking-aware, where we set the Thumb bit from the
> LS bit and r15 from everything else
>  (2) interworking-unaware, where we just set r15 (and it's
> the caller's job to not pass us a misaligned value)
>

Actually I am just going to leave as-is and mask off in the caller.
What I am really trying to do is remove usage of r[15] from boot.c and
I can do that using strategy (2) still.

Regards,
Peter

> This seems to be an odd mix of both.
>
> -- PMM
>