Paul Brook wrote:
> [foreach_cpu]
>>> I tend to dislike such macros...
>> Why?
> 
> I agree with Fabrice in this case. Iterating over all cpus isn't a 
> particularly onerous task to start with. Using bizarre macros means that the 
> code no longer looks like normal C.

OK, got it. Looks like tastes are different here.

> 
>> My debugger SMP fix already comes with 5 more use cases.
> 
> I'm unconvinced by SMP debugger support. FSF gdb isn't currently capable of 
> handling SMP targets, so whatever you're doing is likely to be an ugly hack.

Well, it makes things useable on SMP hosts. Please suggest alternatives
if you think that this is the wrong approach, but leaving this unfixed
would be a mistake IMO.

Also note that gdb used for this case is in no way different from gdb
used with the kgdb stub. In both cases, the debugger (silently) expects
global scope of breakpoints. And that is what my patch ensures + it
enables focus switching for the gdbstub. Again, I'm open for alternative
approaches to achieve comparable usability. I can only repeat that
debugging SMP guest _depends_ on such changes.

Jan