]> git.baikalelectronics.ru Git - kernel.git/commitdiff
mm/page_alloc: fix and rework pfn handling in memmap_init_zone()
authorDavid Hildenbrand <david@redhat.com>
Tue, 4 Feb 2020 01:33:59 +0000 (17:33 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 4 Feb 2020 03:05:23 +0000 (03:05 +0000)
Let's update the pfn manually whenever we continue the loop.  This makes
the code easier to read but also less error prone (and we can directly fix
one issue).

When overlap_memmap_init() returns true, pfn is updated to
"memblock_region_memory_end_pfn(r)".  So it already points at the *next*
pfn to process.  Incrementing the pfn another time is wrong, we might
leave one uninitialized.  I spotted this by inspecting the code, so I have
no idea if this is relevant in practise (with kernelcore=mirror).

Link: http://lkml.kernel.org/r/20200113144035.10848-2-david@redhat.com
Fixes: 31638b96336a ("mm: move mirrored memory specific code outside of memmap_init_zone")
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Baoquan He <bhe@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: "Jin, Zhi" <zhi.jin@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/page_alloc.c

index 7d5b9dbf40874eaa5130a41392d6f992ec39cf3b..461ed73bc30f4ef28f8b4a19c1e44888c3166767 100644 (file)
@@ -5905,18 +5905,20 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
        }
 #endif
 
-       for (pfn = start_pfn; pfn < end_pfn; pfn++) {
+       for (pfn = start_pfn; pfn < end_pfn; ) {
                /*
                 * There can be holes in boot-time mem_map[]s handed to this
                 * function.  They do not exist on hotplugged memory.
                 */
                if (context == MEMMAP_EARLY) {
                        if (!early_pfn_valid(pfn)) {
-                               pfn = next_pfn(pfn) - 1;
+                               pfn = next_pfn(pfn);
                                continue;
                        }
-                       if (!early_pfn_in_nid(pfn, nid))
+                       if (!early_pfn_in_nid(pfn, nid)) {
+                               pfn++;
                                continue;
+                       }
                        if (overlap_memmap_init(zone, &pfn))
                                continue;
                        if (defer_init(nid, pfn, end_pfn))
@@ -5944,6 +5946,7 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
                        set_pageblock_migratetype(page, MIGRATE_MOVABLE);
                        cond_resched();
                }
+               pfn++;
        }
 }