Re: Re : [PATCH] acpi: Fix hot cpu remove problem on acpi subsystem

[Bjorn Helgaas]

On Tue, Sep 6, 2011 at 7:40 PM, canquan.shen <shencanquan@xxxxxxxxxx> wrote:
> On 2011/9/7 2:38, Bjorn Helgaas wrote:

>> Is acpi_processor_remove() called when you remove a processor?  I see
>> a path where it will be called via acpi_eject_store():
>>
>>     acpi_eject_store
>>       acpi_os_hotplug_execute(acpi_bus_hot_remove_device)
>>       acpi_bus_hot_remove_device
>>         acpi_bus_trim
>>           acpi_bus_remove
>>             device_release_driver
>>               dev->driver->remove (acpi_processor_remove())
>>             acpi_device_unregister
>>               device_unregister
>>                 device_del
>>                   kobject_uevent(KOBJ_REMOVE)
>>
>> but as far as I can tell, this path is only used when we write
>> something to the "eject" sysfs file.  I would think we'd want to use
>> most of this same path when we hot remove a CPU via the ACPI SCI
>> mechanism.
>>
>
> Because in my patch will send the KOBJ_REMOVE event to udev module. and I
> write a udev rule like the following:
> ACTION=="remove",DRIVER=="processor",SUBSYSTEM=="acpi",RUN+="/bin/bash -c
> 'echo 1 > /sys%p/eject'"
> This rule will write "1" to the "eject" sysfs file. and then call
> acpi_eject_store function.

Hmmm.  I think I understand your proposal, but it seems like a
convoluted path to me.

I guess the real question is whether we must give userspace a chance
to decide whether to actually do the remove or not.  Is there a
requirement to do that?  Neither the dynamic device removal flow (ACPI
spec 4.0a, sec 6.3) nor the ejection flow example (fig 6-5) mentions
doing that.

I mentioned before that I think the ACPI hotplug code should be ripped
out of the drivers and consolidated in the ACPI core.  I think it's
pretty clear from the spec that the 0-0x7f notifications (Bus Check,
Device Check, Eject Request, etc.) are designed to be handled by the
core, not by individual drivers.  We handle hotplug in the drivers
today, but I think that's mainly because we never implemented support
in the Linux ACPI core.  There are comments in acpi_bus_check_device()
and acpi_bus_check_scope() about what we *should* be doing there.

I am opposed to adding more hotplug support to individual drivers
because I still hope that someday we'll support it in the ACPI core.
Many ACPI drivers don't support hotplug at all, and the ones that do
support hotplug do it in a variety of ways.  It's all quite a mess.

Bjorn
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/