Messages in this thread |  | | Date | Fri, 17 Jul 2009 15:20:53 +0900 | From | Yasunori Goto <> | Subject | Re: [RFC patch] delete improper hot pluggable code of memory affinity |
| |
> On Fri, Jul 17, 2009 at 1:52 PM, Yasunori Goto<y-goto@jp.fujitsu.com> wrote: > >> Without a fix like my proposal, I have seen NUMA configure disabled by > >> kernel (due to the code the patch deletes) on a system with Enabled > >> bit set , and Hotplug-able bit cleared, and > >> CONFIG_MEMORY_HOTPLUG_SPARSE disabled. > > > > Ok. I guess that save_add_info() was to check percentage of > > reserve area when CONFIG_MEMORY_HOTPLUG_RESERVE is set. > > Its code was removed at 2.6.25, save_add_info() may be garbage now. > > > > However, I have one question now. > > > >>> - ? ?if (ma->flags & ACPI_SRAT_MEM_HOT_PLUGGABLE) { > >>> - ? ? ? ? ? ?update_nodes_add(node, start, end); > >>> - ? ? ? ? ? ?/* restore nodes[node] */ > >>> - ? ? ? ? ? ?*nd = oldnode; > >>> - ? ? ? ? ? ?if ((nd->start | nd->end) == 0) > >>> - ? ? ? ? ? ? ? ? ? ?node_clear(node, nodes_parsed); > >>> - ? ? } > > > > I don't understand why you remove this code. could you tell me why? > > Good question. This is exactly the place I'm puzzled too. > Without delete this code, I still see one fake node instead of 4 real node... > I think a flow up patch is needed here...
In my understanding, here is for the case when the SRAT entry is enabled, but memory is not connected. (This is why I sent the first mail). However, here may have a bug yet.
I would like to test here, but I don't have a real x86-64 machine which can hot add memory. So, I can just guess now.
Anyway, thanks for your notification. If you have progress about it, please let me know.
Bye.
-- Yasunori Goto
|  |