mm/page_owner: don't access uninitialized memmaps when reading /proc/pagetypeinfo
authorQian Cai <cai@lca.pw>
Sat, 19 Oct 2019 03:19:29 +0000 (20:19 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 19 Oct 2019 10:32:31 +0000 (06:32 -0400)
Uninitialized memmaps contain garbage and in the worst case trigger
kernel BUGs, especially with CONFIG_PAGE_POISONING.  They should not get
touched.

For example, when not onlining a memory block that is spanned by a zone
and reading /proc/pagetypeinfo with CONFIG_DEBUG_VM_PGFLAGS and
CONFIG_PAGE_POISONING, we can trigger a kernel BUG:

  :/# echo 1 > /sys/devices/system/memory/memory40/online
  :/# echo 1 > /sys/devices/system/memory/memory42/online
  :/# cat /proc/pagetypeinfo > test.file
   page:fffff2c585200000 is uninitialized and poisoned
   raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
   raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
   page dumped because: VM_BUG_ON_PAGE(PagePoisoned(p))
   There is not page extension available.
   ------------[ cut here ]------------
   kernel BUG at include/linux/mm.h:1107!
   invalid opcode: 0000 [#1] SMP NOPTI

Please note that this change does not affect ZONE_DEVICE, because
pagetypeinfo_showmixedcount_print() is called from
mm/vmstat.c:pagetypeinfo_showmixedcount() only for populated zones, and
ZONE_DEVICE is never populated (zone->present_pages always 0).

[david@redhat.com: move check to outer loop, add comment, rephrase description]
Link: http://lkml.kernel.org/r/20191011140638.8160-1-david@redhat.com
Fixes: f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online") # visible after d0dc12e86b319
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Miles Chen <miles.chen@mediatek.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: <stable@vger.kernel.org> [4.13+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/page_owner.c

index e327bcd0380e9d79a5f974c3a4f56889e026cd0f..18ecde9f45b24c069172045ef2f829fc94a1133f 100644 (file)
@@ -271,7 +271,8 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
         * not matter as the mixed block count will still be correct
         */
        for (; pfn < end_pfn; ) {
-               if (!pfn_valid(pfn)) {
+               page = pfn_to_online_page(pfn);
+               if (!page) {
                        pfn = ALIGN(pfn + 1, MAX_ORDER_NR_PAGES);
                        continue;
                }
@@ -279,13 +280,13 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
                block_end_pfn = ALIGN(pfn + 1, pageblock_nr_pages);
                block_end_pfn = min(block_end_pfn, end_pfn);
 
-               page = pfn_to_page(pfn);
                pageblock_mt = get_pageblock_migratetype(page);
 
                for (; pfn < block_end_pfn; pfn++) {
                        if (!pfn_valid_within(pfn))
                                continue;
 
+                       /* The pageblock is online, no need to recheck. */
                        page = pfn_to_page(pfn);
 
                        if (page_zone(page) != zone)