Re: [Qemu-devel] [PATCH] v3: don't call reset functions on cpu initialization

[Blue Swirl <blauwirbel@gmail.com>]

On Fri, Nov 6, 2009 at 10:41 PM, Glauber Costa <glommer@redhat.com> wrote:
> On Fri, Nov 06, 2009 at 09:01:45PM +0200, Blue Swirl wrote:
>> On Fri, Nov 6, 2009 at 8:43 PM, Laurent Desnogues
>> <laurent.desnogues@gmail.com> wrote:
>> > On Fri, Nov 6, 2009 at 7:11 PM, Glauber Costa <glommer@redhat.com> wrote:
>> >> On Fri, Nov 06, 2009 at 08:05:40PM +0200, Blue Swirl wrote:
>> >>> On Thu, Nov 5, 2009 at 6:05 PM, Glauber Costa <glommer@redhat.com> wrote:
>> > [...]
>> >>> > --- a/target-i386/helper.c
>> >>> > +++ b/target-i386/helper.c
>> >>> > @@ -1885,7 +1885,9 @@ CPUX86State *cpu_x86_init(const char *cpu_model)
>> >>> >         return NULL;
>> >>> >     }
>> >>> >     mce_init(env);
>> >>> > +#ifdef CONFIG_USER_ONLY
>> >>> >     cpu_reset(env);
>> >>> > +#endif
>> >>>
>> >>> Please push the call to *-user/main.c, just after call to cpu_init().
>> >> I'd prefer it that way too. But cpu_reset is also called in some other places,
>> >> and Laurent suggested me to to this way.
>> >
>> > Yes, you'd have to remove all calls to cpu_reset from all CPUs.
>> > And also add a call to cpu_reset to cpu_copy.
>> >
>> >> I don't really know much about -user, so I'm fine with whatever you guys agree on.
>> >
>> > I honestly don't care that much as long as all targets still work
>> > in user mode :-)
>> >
>> > The aim was to make Glauber's patch less intrusive.
>>
>> Given that only the new calls to cpu_reset are important and the
>> removals are much less so (double reset shouldn't be a problem), the
>> least intrusive version would be to just add the new calls and do the
>> clean up later.
> Which IMHO, pretty much means apply it this way, and then later on move
> the reset functions elsewhere for -user, if one wants to.

Thanks, applied. While testing I noticed that the poor x86 emulator
does not survive even five system_resets until BIOS refuses to boot.
This happens without your patch. In comparison, I see no problems
resetting Sparc32, Sparc64 or PPC.