Re: [PATCH v2 2/2] Do not change worker's running cpu in cmci_rediscover().

[Tang Chen]

On 10/22/2012 06:18 PM, Borislav Petkov wrote:
> On Mon, Oct 22, 2012 at 11:33:16AM +0800, Tang Chen wrote:
> > I have 2 nodes, node0 and node1. node1 could be hotpluged.
> > node0 has cpu0 ~ cpu15, node1 has cpu16 ~ cpu31.
> >
> > I online all the cpus on node1, and hot-remove node1 directly.
>
> Hold on, I need to ask here: you soft-online all cores on node1 and
> *then* you *hot* *remove* it? So with all cores online you physically
> take out the processor from the socket? Am I reading this correctly?

Hi Borislav,

No, it's not like that. I'm sorry to make you confused.

Firstly, let me do a little explanation here. :)

I was doing ACPI based hotplug. In a container device, it contains
cpus and memories and so on. I didn't physically remove cpus from
hardware. I emulated a SCI with my own test module, and the
container_notify_cb() was called, and it recursively remove all the
sub-devices.

The "remove" here doesn't mean "taking it away". And of course, when
the "remove" is done, you can physically take the whole system board
away.

Please refer to this url:
http://www.spinics.net/lists/linux-pci/msg17651.html
This is the module we are doing the SCI emulation.

In summary, the hotplug process could be like the following:

hot-add(SCI) -> online(softly) -> device working -> offline(softly) -> 
hot-remove(SCI)


Secondly, in kernel, before a container device is removed, all its
sub-devices will be offlined and removed first. So I have all the cpus
online, and hot-remove a container(which I think is a node, maybe not)
directly. This operation has no problem, I think.

Maybe I should not say container as a node. :)

And again, sorry for the confusion.

Thanks. :)


> Thanks.

--
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/