[Qemu-devel] Re: [PATCH 02/17] lm32: translation routines

[Alexander Graf]

On 07.02.2011, at 23:55, Michael Walle wrote:

> On Mon, Feb 07 2011, 23:20:33, Edgar E. Iglesias wrote:
>> Yes, I found the nop tracking handy too but It exposes a way for guests
>> to completely abort the vm with valid code. On microblaze, even guest
>> userspace can trig the cpu_abort. Your port doesn't have an MMU so guest
>> userland can probably kill the vm anyway, so it might be less of an issue.
>> 
>> Maybe the event could be logged instead of cpu_abort:ed?
> If the program jumps to uninitialized memory, it'll likely execute until the 
> end of the memory and then kill the VM.
> Maybe we can add some kind of runtime configuration option for 
> enabling/disabling this?

During the first few reviews I stumbled over this too and actually found it a 
pretty slick hack. I agree though that it'd be nice to have a machine switch. 
Don't we have some machine config options these days? Alternatively, it could 
even be a -cpu flag.


Alex