Re: [PATCH] x86, apic: clean up handling of boot_cpu_physical_apicid in boot process

[HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>]

(2014/01/26 13:02), David Rientjes wrote:
> On Thu, 16 Jan 2014, HATAYAMA Daisuke wrote:
> 
>> Hello,
>>
>> This patch deals with the issue of handling boot_cpu_physical_apicid
>> in boot process I avoided in disable_cpu_apicid patch because I
>> cannot guess how long it needs to take for the review of this fix.
>>
>> This patch is made on top of today's x86/apic branch of tip tree.
>> Its commit hash is 5b4d1dbc24bb6fd7179ada0f47be34e27e64decb
>>
> 
> This breaks with SGI 320/540:
> 
> arch/x86/platform/visws/visws_quirks.c: In function ‘MP_processor_info’:
> arch/x86/platform/visws/visws_quirks.c:169:3: error: ‘bios_cpu_physical_apicid’ undeclared (first use in this function)
> arch/x86/platform/visws/visws_quirks.c:169:3: note: each undeclared identifier is reported only once for each function it appears in
> 

Thanks. I'll retry with CONFIG_X86_VISWS...

obj-$(CONFIG_X86_VISWS) += visws_quirks.o

> It makes it pretty apparent that you want a different name for the new
> variable, bios_bsp_physical_apicid is just too close to
> boot_cpu_physical_apicid that even the author of the patch missed a
> conversion.
> 

``bsp'' and ``boot cpu'' are different, but certainly they are close
for most of people who don't need to focus the difference.

One idea is not to add bios_bsp_physical_apicid, just removing
assignment to boot_cpu_physical_apicid from MP_processor_info().
I guess currently no one uses apicid provided by MP table
since boot_cpu_physical_apicid is finally initialized in
init_apic_mappings().

> (It's also really sad that we can't make it __initdata since it's only
> valid at boot because of the hotplug stuff.)
> 


-- 
Thanks.
HATAYAMA, Daisuke