[PATCH 1/2] x86/kasan: clear kasan_zero_page after TLB flush

From: Andrey Ryabinin
Date: Mon Jan 11 2016 - 07:50:57 EST


Currently we clear kasan_zero_page before __flush_tlb_all(). This
works with current implementation of native_flush_tlb[_global]()
because it doesn't cause do any writes to kasan shadow memory.
But any subtle change made in native_flush_tlb*() could break this.
Also current code seems doesn't work for paravirt guests (lguest).

Only after the TLB flush we can be sure that kasan_zero_page is not
used as early shadow anymore (instrumented code will not write to it).
So it should cleared it only after the TLB flush.

Signed-off-by: Andrey Ryabinin <aryabinin@xxxxxxxxxxxxx>
---
arch/x86/mm/kasan_init_64.c | 11 ++++++++---
1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/arch/x86/mm/kasan_init_64.c b/arch/x86/mm/kasan_init_64.c
index d470cf2..303e470 100644
--- a/arch/x86/mm/kasan_init_64.c
+++ b/arch/x86/mm/kasan_init_64.c
@@ -120,11 +120,16 @@ void __init kasan_init(void)
kasan_populate_zero_shadow(kasan_mem_to_shadow((void *)MODULES_END),
(void *)KASAN_SHADOW_END);

- memset(kasan_zero_page, 0, PAGE_SIZE);
-
load_cr3(init_level4_pgt);
__flush_tlb_all();
- init_task.kasan_depth = 0;

+ /*
+ * kasan_zero_page has been used as early shadow memory, thus it may
+ * contain some garbage. Now we can clear it, since after the TLB flush
+ * no one should write to it.
+ */
+ memset(kasan_zero_page, 0, PAGE_SIZE);
+
+ init_task.kasan_depth = 0;
pr_info("KernelAddressSanitizer initialized\n");
}
--
2.4.10