Re: [Qemu-devel] Buggy wfi support for ARM user

[Laurent Desnogues]

On Sun, Jun 8, 2008 at 6:38 PM, Blue Swirl <address@hidden> wrote:
> On 6/8/08, Laurent Desnogues <address@hidden> wrote:
>> >> I propose that for user mode this instruction be considered
>>  >> as a NOP.  Any thoughts?
>>  >
>>  > Allowing usermode to issue WFI sounds like a bug.
>>
>>  Here is a proposal that makes WFI no-ops for user mode emulation.
>>  Comments welcome, especially as I am not used to send patches :)
>
> Generally, the config-host.mak part should be suppressed in patches.

Sorry about that, I used make clean instead of distclean.

> I don't know about this WFI case or even much of ARM at all, but if
> enabling WFI is controversial, maybe it could be implemented as a CPU
> feature so that it is disabled by default.

It's not controversial I think.  It's just that it's architected to be
executable by user processes.  However it only makes sense
if a process can receive interruptions which is not really the
case for qemu user mode emulation.  This instruction is used
to reduce power in idle loops.


Laurent

wfi.patch
Description: Text Data