Re: [Lhms-devel] [RFC/PATCH: 002/010] Memory hotplug for new nodes with pgdat allocation. (Wait table and zonelists initalization)

[Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

Yasunori Goto wrote:
> 
>> On Fri, 2006-02-10 at 23:20 +0900, Yasunori Goto wrote:
>>>  static __meminit
>>>  void zone_wait_table_init(struct zone *zone, unsigned long
>>> zone_size_pages)
>>>  {
>>> -       int i;
>>> +       int i, hotadd = (system_state == SYSTEM_RUNNING);
>>>         struct pglist_data *pgdat = zone->zone_pgdat;
>>> +       unsigned long allocsize;
>>>  
>>>         /*
>>>          * The per-page waitqueue mechanism uses hashed waitqueues
>>>          * per zone.
>>>          */
>>> +       if (hotadd && (zone_size_pages == PAGES_PER_SECTION))
>>> +               zone_size_pages = PAGES_PER_SECTION << 2; 
>> I don't think I understand this calculation.  You online only 4 sections
>> worth of pages?
> 
> Ummmmm.
> I realized that I've forgotten many things about this patch
> due to long time keeping in storage. 
> At least here looks strange indeed.
> I need shake my brain to recall it. :-(

Ah, I'm not sure but it was because I didn't have a patch for zone's waittable resizing,
and resizing it looked impossible. Above code was just a quick hack for the case a zone
is initialized with only 1 section.
How large it should be ? or Resizing it, is necessary to be discussed.

-- Kame