Re: [PATCH 2.6][resend] Add DEVPATH env variable to hotplug helper call

[Keshavamurthy Anil S <anil.s.keshavamurthy@intel.com>]

On Mon, Oct 04, 2004 at 12:37:25PM -0700, Andrew Morton wrote:
> Keshavamurthy Anil S <anil.s.keshavamurthy@intel.com> wrote:
> >
> >  On Sun, Oct 03, 2004 at 04:20:12PM -0700, Andrew Morton wrote:
> >  > Does CPU hotplug behave correctly wrt /sys/devices/system/cpu?  Given that
> >  > register_cpu() is still marked __init, I assume not.
> > 
> >  Currently what we have in the kernel is logical cpu hotplug, i.e once the
> >  cpu is registered via register_cpu() that cpu can only go offline and still
> >  the entry for that cpu will be present in the /sys/devices/system/cpu/cpuX/online.
> > 
> >  So __init register_cpu() is fine untill we support unregister_cpu()
> >  which is required for physical cpu hotplug case.
> > 
> >  I have submitted ACPI based physical cpu hotplug patches and waiting to here from
> >  ACPI mainitainer Len Brown, there I have taken care to support unregister_cpu()
> >  and register_cpu() is marked as __devinit in those patches.
> 
> OK...
> 
> But still, cpu_run_sbin_hotplug() should not exist.  It is duplicating
> (indeed, emulating) kobject_hotplug() behaviour.  To the extent that it now
> has a hardwired sysfs path embedded in it:
> 
> 	sprintf(devpath_str, "DEVPATH=devices/system/cpu/cpu%d", cpu);
> 
> which should have been obtained from kobject_get_path().
Yes, I agree to your point that cpu_run_sbin_hotplug() is duplication kobject_hotplug()
behaviour. I will send you a patch to fix this ASAP(hopefully before end of today).

thanks,
Anil