[PATCH AUTOSEL 5.0 044/262] page_poison: play nicely with KASAN

[Sasha Levin <sashal@kernel.org>]

From: Qian Cai <cai@lca.pw>

[ Upstream commit 4117992df66a26fa33908b4969e04801534baab1 ]

KASAN does not play well with the page poisoning (CONFIG_PAGE_POISONING).
It triggers false positives in the allocation path:

  BUG: KASAN: use-after-free in memchr_inv+0x2ea/0x330
  Read of size 8 at addr ffff88881f800000 by task swapper/0
  CPU: 0 PID: 0 Comm: swapper Not tainted 5.0.0-rc1+ #54
  Call Trace:
   dump_stack+0xe0/0x19a
   print_address_description.cold.2+0x9/0x28b
   kasan_report.cold.3+0x7a/0xb5
   __asan_report_load8_noabort+0x19/0x20
   memchr_inv+0x2ea/0x330
   kernel_poison_pages+0x103/0x3d5
   get_page_from_freelist+0x15e7/0x4d90

because KASAN has not yet unpoisoned the shadow page for allocation
before it checks memchr_inv() but only found a stale poison pattern.

Also, false positives in free path,

  BUG: KASAN: slab-out-of-bounds in kernel_poison_pages+0x29e/0x3d5
  Write of size 4096 at addr ffff8888112cc000 by task swapper/0/1
  CPU: 5 PID: 1 Comm: swapper/0 Not tainted 5.0.0-rc1+ #55
  Call Trace:
   dump_stack+0xe0/0x19a
   print_address_description.cold.2+0x9/0x28b
   kasan_report.cold.3+0x7a/0xb5
   check_memory_region+0x22d/0x250
   memset+0x28/0x40
   kernel_poison_pages+0x29e/0x3d5
   __free_pages_ok+0x75f/0x13e0

due to KASAN adds poisoned redzones around slab objects, but the page
poisoning needs to poison the whole page.

Link: http://lkml.kernel.org/r/20190114233405.67843-1-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Acked-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 mm/page_alloc.c  | 2 +-
 mm/page_poison.c | 4 ++++
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 0b9f577b1a2a..10d0f2ed9f69 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1945,8 +1945,8 @@ inline void post_alloc_hook(struct page *page, unsigned int order,
 
 	arch_alloc_page(page, order);
 	kernel_map_pages(page, 1 << order, 1);
-	kernel_poison_pages(page, 1 << order, 1);
 	kasan_alloc_pages(page, order);
+	kernel_poison_pages(page, 1 << order, 1);
 	set_page_owner(page, order, gfp_flags);
 }
 
diff --git a/mm/page_poison.c b/mm/page_poison.c
index f0c15e9017c0..21d4f97cb49b 100644
--- a/mm/page_poison.c
+++ b/mm/page_poison.c
@@ -6,6 +6,7 @@
 #include <linux/page_ext.h>
 #include <linux/poison.h>
 #include <linux/ratelimit.h>
+#include <linux/kasan.h>
 
 static bool want_page_poisoning __read_mostly;
 
@@ -40,7 +41,10 @@ static void poison_page(struct page *page)
 {
 	void *addr = kmap_atomic(page);
 
+	/* KASAN still think the page is in-use, so skip it. */
+	kasan_disable_current();
 	memset(addr, PAGE_POISON, PAGE_SIZE);
+	kasan_enable_current();
 	kunmap_atomic(addr);
 }
 
-- 
2.19.1