On Fri, Jun 16, 2017 at 10:33:50AM -0700, Andrew Morton wrote:
>On Fri, 16 Jun 2017 17:23:35 +0800 Wei Yang <richard.weiyang@xxxxxxxxx> wrote:
>
>> In function move_pfn_range_to_zone(), memmap_init_zone() will call
>> set_page_links for each page.
>
>Well, no. There are several types of pfn's for which
>memmap_init_zone() will not call
>__init_single_page()->set_page_links(). Probably the code is OK, as
>those are pretty screwy pfn types. But I'd like to see some
>confirmation that this patch is OK for all such pfns, now and in the
>future?
>
Hmm... when memmap_init_zone() is called during hotplug, this means
(context != MEMMAP_EARLY). So it will jump to the end and call
__init_single_page().
Is my understanding corrent?
>> This means we don't need to call it on each
>> page explicitly.
>>
>> This patch just removes the loop.
>>
>> ...
>>
>> --- a/mm/memory_hotplug.c
>> +++ b/mm/memory_hotplug.c
>> @@ -914,10 +914,6 @@ void __ref move_pfn_range_to_zone(struct zone *zone,
>> * are reserved so nobody should be touching them so we should be safe
>> */
>> memmap_init_zone(nr_pages, nid, zone_idx(zone), start_pfn, MEMMAP_HOTPLUG);
>> - for (i = 0; i < nr_pages; i++) {
>> - unsigned long pfn = start_pfn + i;
>> - set_page_links(pfn_to_page(pfn), zone_idx(zone), nid, pfn);
>> - }
>>
>> set_zone_contiguous(zone);
>> }
--
Wei Yang
Help you, Help me
Attachment:
signature.asc
Description: PGP signature
