Re: [Qemu-devel] [PATCH v2 1/2] hmp: dump ids including socket-id, core-id and so on for 'info registers'

[Eduardo Habkost <ehabkost@redhat.com>]

On Thu, Jul 27, 2017 at 05:12:45PM +0200, Igor Mammedov wrote:
> On Wed, 26 Jul 2017 14:50:10 -0300
> Eduardo Habkost <ehabkost@redhat.com> wrote:
> 
> > On Wed, Jul 26, 2017 at 04:27:54PM +0200, Igor Mammedov wrote:
> > > On Wed, 26 Jul 2017 11:01:14 -0300
> > > Eduardo Habkost <ehabkost@redhat.com> wrote:
> > >   
> > > > On Wed, Jul 26, 2017 at 02:18:36AM -0400, Yi Wang wrote:  
> > > > > This patch add output of CPUs' socket-id, core-id, thread-id and
> > > > > apic-id for 'info registers', which can be used for querying other
> > > > > hmp commands.
> > > > > 
> > > > > Signed-off-by: Yi Wang <wang.yi59@zte.com.cn>
> > > > > Signed-off-by: Yun Liu <liu.yunh@zte.com.cn>
> > > > > ---
> > > > >  target/i386/helper.c | 8 ++++++++
> > > > >  1 file changed, 8 insertions(+)
> > > > > 
> > > > > diff --git a/target/i386/helper.c b/target/i386/helper.c
> > > > > index f63eb3d..a52f300 100644
> > > > > --- a/target/i386/helper.c
> > > > > +++ b/target/i386/helper.c
> > > > > @@ -416,6 +416,14 @@ void x86_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
> > > > >      int eflags, i, nb;
> > > > >      char cc_op_name[32];
> > > > >      static const char *seg_name[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
> > > > > +    APICCommonState *s = APIC_COMMON(cpu->apic_state);
> > > > > +
> > > > > +    if (!s) {
> > > > > +        cpu_fprintf(f, "local apic state not available\n");
> > > > > +        return;
> > > > > +    }   
> > > > 
> > > > This breaks "info registers" on 486:
> > > > 
> > > >   $ ./x86_64-softmmu/qemu-system-x86_64 -cpu 486 -display none -monitor stdio
> > > >   QEMU 2.9.90 monitor - type 'help' for more information
> > > >   (qemu) info registers
> > > >   local apic state not available
> > > >   (qemu)
> > > > 
> > > > You can simply use X86CPU::apic_id.  The field is set to a consistent value
> > > > even if there's no guest-visible APIC, because we use it as a VCPU identifier
> > > > internally.
> > > > 
> > > > However, I'm not sure "info registers" is the right place to print this info.
> > > > I think "info cpus" would be a more obvious place, but it would probably make
> > > > its output too long.
> > > > We could implement something like "info cpus -v", but: the "apic-id" info would
> > > > be redundant as soon as we replace "CPU #<cpu_index>" with "CPU #<arch_id>",  
> > > lets do it in 'info registers' for now, it looks like obvious place to display
> > > yet another cpu register.  
> > 
> > I think the APIC ID is useful to be printed by "info cpus" and
> > "info registers", but socket/core/thread IDs don't belong to
> > "info registers" and can be printed by "info qtree".
> > 
> > But I don't see a point in adding APIC ID temporarily to "info
> > registers" if it will be made redundant before the next release
> > (hopefully) when we start using arch_id on the "CPU #%d" lines.
> > (It is already too late for 2.10.)
> we can't switch to arch_id unless it's done for all related commands in one go.

That's what I'm proposing: switching all commands to use arch_id
to identify CPUs.

> 
> > 
> > > 
> > >   
> > > > and other details like socket/core/thread could just appear on "info qtree".  
> > > cpus are not displayed in 'info qtree'  
> > 
> > We can fix that.
> qtree is based on qdev bus hierarchy and CPUs (at least for i386 are bus-less),
> not sure I'd like to back and tie them to some 'bus' again

We can fix that by making info qtree print bus-less devices too.

-- 
Eduardo