mm: use early_pfn_to_nid in page_ext_init

page_ext_init() checks suitable pages with pfn_to_nid(), but
pfn_to_nid() depends on memmap which will not be setup fully until
page_alloc_init_late() is done.  Use early_pfn_to_nid() instead of
pfn_to_nid() so that page extension could be still used early even
though CONFIG_ DEFERRED_STRUCT_PAGE_INIT is enabled and catch early page
allocation call sites.

Suggested by Joonsoo Kim [1], this fix basically undoes the change
introduced by commit b8f1a75d61d840 ("mm: call page_ext_init() after all
struct pages are initialized") and fixes the same problem with a better
approach.

[1] http://lkml.kernel.org/r/CAAmzW4OUmyPwQjvd7QUfc6W1Aic__TyAuH80MLRZNMxKy0-wPQ@mail.gmail.com

Link: http://lkml.kernel.org/r/1464198689-23458-1-git-send-email-yang.shi@linaro.org
Signed-off-by: Yang Shi <yang.shi@linaro.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 1 insertions, 2 deletions

diff --git a/init/main.c b/init/main.c
index bc0f9e0bcf22..4c17fda5c2ff 100644
--- a/init/main.c
+++ b/init/main.c
@@ -607,6 +607,7 @@ asmlinkage __visible void __init start_kernel(void)
                 initrd_start = 0;
         }
 #endif
+        page_ext_init();
         debug_objects_mem_init();
         kmemleak_init();
         setup_per_cpu_pageset();
@@ -1003,8 +1004,6 @@ static noinline void __init kernel_init_freeable(void)
         sched_init_smp();
 
         page_alloc_init_late();
-        /* Initialize page ext after all struct pages are initializaed */
-        page_ext_init();
 
         do_basic_setup();