Re: [PATCH] x86: read apic id if it is not acpi_lapic -v2

[Ingo Molnar <mingo@elte.hu>]

* Cyrill Gorcunov <gorcunov@gmail.com> wrote:

> [Ingo Molnar - Mon, May 11, 2009 at 11:26:55AM +0200]
> | 
> | * Yinghai Lu <yinghai@kernel.org> wrote:
> | 
> | > @@ -1590,21 +1591,32 @@ void __init init_apic_mappings(void)
> | >  	} else
> | >  		apic_phys = mp_lapic_addr;
> | >  
> | > -	set_fixmap_nocache(FIX_APIC_BASE, apic_phys);
> | > +	/* lets check if we may to NOP'ify apic operations */
> | > +	if (!cpu_has_apic) {
> | > +		pr_info("APIC: disable apic facility\n");
> | > +		apic_disable();
> | > +		return;
> | > +	}
> | > +
> | > +	/*
> | > +	 * acpi lapic path already map that address in
> | > +	 * acpi_register_lapic_address()
> | > +	 */
> | >  	apic_printk(APIC_VERBOSE, "mapped APIC to %08lx (%08lx)\n",
> | >  				APIC_BASE, apic_phys);
> | > +	if (acpi_lapic)
> | > +		return;
> | 
> | why do we return here? We should sanity-check the APIC ID even if we 
> | come from ACPI.
> | 
> | 	Ingo
> | 
> 
> I believe we already have it checked in 
> acpi_register_lapic_address so no need to do it twice. Or I miss 
> something?

It's better to have such checks applied redundantly instead of 
introducing such implicit criss-cross dependencies between 
functions. (which would break if acpi_register_lapic_address() ever 
stopped checking the apic id - or did it differently)

	Ingo