Re: More than 8 CPUs detected and CONFIG_X86_PC cannot handle it on 2.6.16-rc6

[Andrew Morton <akpm@osdl.org>]

Ashok Raj <ashok.raj@intel.com> wrote:
>
> 
> 
>  When CONFIG_HOTPLUG_CPU is turned on we always use physflat mode (bigsmp) even 
>  when #of CPUs are less than 8 to avoid sending IPI to offline processors.
> 
>  Without having BIGSMP on it spits out a warning during boot on systems that
>  seems misleading, since it complains even on systems that have less
>  than 8 cpus.
> 
>  Signed-off-by: Ashok Raj <ashok.raj@intel.com>
>  ---------------------------------------------------------
> 
>   arch/i386/Kconfig |    2 +-
>   1 files changed, 1 insertion(+), 1 deletion(-)
> 
>  Index: linux-2.6.16-rc6-mm1/arch/i386/Kconfig
>  ===================================================================
>  --- linux-2.6.16-rc6-mm1.orig/arch/i386/Kconfig
>  +++ linux-2.6.16-rc6-mm1/arch/i386/Kconfig
>  @@ -760,7 +760,7 @@ config PHYSICAL_START
>   
>   config HOTPLUG_CPU
>   	bool "Support for hot-pluggable CPUs (EXPERIMENTAL)"
>  -	depends on SMP && HOTPLUG && EXPERIMENTAL && !X86_VOYAGER
>  +	depends on SMP && HOTPLUG && EXPERIMENTAL && !X86_VOYAGER && !X86_PC
>   	---help---
>   	  Say Y here to experiment with turning CPUs off and on.  CPUs
>   	  can be controlled through /sys/devices/system/cpu.

Still seems wrong.  People _do_ use HOTPLUG_CPU on X86_PCs so they can get
software suspend.  The number of people who do this are probably 100000x
the number of people who have physically hotpluggable CPUs.  And I don't
think we can churn their config requirements this much so late in the game.

So for now I suggest we're best off simply killing the printk (or doing
something smarter, like comparing cpu_online-map with cpu_possible_map
(which isn't right)).

Longer term, it appears that we need to do some Kconfig and C work to
separate out the HOTPLUG_CPU infrastructure which swsusp needs from actual
CPU hotplugging.

What _is_ this IPI problem anyway?  Can't send point-to-point IPIs to
offlined CPUs?  (Don't do that then?) Or do broadcast IPIs go wrong, or
what?

And does it affect pretend-x86-hotplug, or is it only affecting real hotplug?

Thanks.