Re: [tip:irq/numa] x86: read apic ID in the !acpi_lapic case

[Cyrill Gorcunov <gorcunov@gmail.com>]

[Yinghai Lu - Tue, May 12, 2009 at 10:12:05AM -0700]
| Cyrill Gorcunov wrote:
| > [Ingo Molnar - Tue, May 12, 2009 at 06:46:17PM +0200]
| > | 
| > | got this on a testbox:
| > | 
| > | [    0.113333] WARNING: at arch/x86/kernel/apic/apic.c:253 warn_slowpath_null+0x28/0x50()
| > | 
| > | 	Ingo
| > |
| > 
| > It's expected (unfortunately). If we have SMP compiled kernel
| > we still rely on apic_read heavily. Since for fake'ed apic
| > we just return 0 for any reading I suspect we could just drop
| > WARN_ON_ONCE for native_apic_read_dummy. Yinghai?
| > 
| > (well, actually we should clean up callees but it would require
| >  some effort)
| 
| 32bit seems doesn't have this warning.
| 
| and on 64bit, we always assumed have local APIC there. so could be that
| noapic is not cleanly enforced
|

yes, that is why I said that we need to cleanup the call-pathes.
And since it requires some effort we either could save WARN_ON_ONCE
here (so it hit us everyday provoking to do cleanup) and keep
in mind that it's expected behaviour at moment, or if it's that
annoying -- remove it (though it would be just a hidding of code
defect I think). Unfortunatelly I have a number of tasks to be done
before I could start such a checking/cleanups.

| 
| YH
| 
	-- Cyrill