From 1ae7013dfa4281644a9a9591105c24d639216b22 Mon Sep 17 00:00:00 2001 From: Hui Zhu Date: Thu, 14 May 2015 15:17:04 -0700 Subject: [PATCH] CMA: page_isolation: check buddy before accessing it I had an issue: Unable to handle kernel NULL pointer dereference at virtual address 0000082a pgd = cc970000 [0000082a] *pgd=00000000 Internal error: Oops: 5 [#1] PREEMPT SMP ARM PC is at get_pageblock_flags_group+0x5c/0xb0 LR is at unset_migratetype_isolate+0x148/0x1b0 pc : [] lr : [] psr: 80000093 sp : c7029d00 ip : 00000105 fp : c7029d1c r10: 00000001 r9 : 0000000a r8 : 00000004 r7 : 60000013 r6 : 000000a4 r5 : c0a357e4 r4 : 00000000 r3 : 00000826 r2 : 00000002 r1 : 00000000 r0 : 0000003f Flags: Nzcv IRQs off FIQs on Mode SVC_32 ISA ARM Segment user Control: 10c5387d Table: 2cb7006a DAC: 00000015 Backtrace: get_pageblock_flags_group+0x0/0xb0 unset_migratetype_isolate+0x0/0x1b0 undo_isolate_page_range+0x0/0xdc __alloc_contig_range+0x0/0x34c alloc_contig_range+0x0/0x18 This issue is because when calling unset_migratetype_isolate() to unset a part of CMA memory, it try to access the buddy page to get its status: if (order >= pageblock_order) { page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1); buddy_idx = __find_buddy_index(page_idx, order); buddy = page + (buddy_idx - page_idx); if (!is_migrate_isolate_page(buddy)) { But the begin addr of this part of CMA memory is very close to a part of memory that is reserved at boot time (not in buddy system). So add a check before accessing it. [akpm@linux-foundation.org: use conventional code layout] Signed-off-by: Hui Zhu Suggested-by: Laura Abbott Suggested-by: Joonsoo Kim Cc: Vlastimil Babka Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- mm/page_isolation.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/mm/page_isolation.c b/mm/page_isolation.c index 755a42c76eb4..303c908790ef 100644 --- a/mm/page_isolation.c +++ b/mm/page_isolation.c @@ -101,7 +101,8 @@ void unset_migratetype_isolate(struct page *page, unsigned migratetype) buddy_idx = __find_buddy_index(page_idx, order); buddy = page + (buddy_idx - page_idx); - if (!is_migrate_isolate_page(buddy)) { + if (pfn_valid_within(page_to_pfn(buddy)) && + !is_migrate_isolate_page(buddy)) { __isolate_free_page(page, order); kernel_map_pages(page, (1 << order), 1); set_page_refcounted(page);