Re: [PATCH] x86/alternatives: Disable KASAN on text_poke_early() in apply_alternatives()

[Kirill A. Shutemov]

On Tue, Oct 10, 2023 at 11:12:35AM +0200, Borislav Petkov wrote:
> On Tue, Oct 10, 2023 at 11:40:41AM +0300, Kirill A. Shutemov wrote:
> > __VIRTUAL_MASK_SHIFT used in many places. I don't think it is good idea to
> > give up on patching completely.
> 
> Have you even looked at boot_cpu_has()'s asm?

Obviously not :/

Okay, as alternative, the patch below also make the issue go away.

But I am not sure it is fundamentaly better. boot_cpu_has() generates call
to __asan_load8_noabort(). I think it only works because all KASAN code
has ASAN instrumentation disabled.

diff --git a/arch/x86/include/asm/kasan.h b/arch/x86/include/asm/kasan.h
index de75306b932e..bfe97013abb0 100644
--- a/arch/x86/include/asm/kasan.h
+++ b/arch/x86/include/asm/kasan.h
@@ -12,8 +12,15 @@
  * for kernel really starts from compiler's shadow offset +
  * 'kernel address space start' >> KASAN_SHADOW_SCALE_SHIFT
  */
+
+#ifdef USE_EARLY_PGTABLE_L5
+#define __KASAN_VIRT_SHIFT	(__pgtable_l5_enabled ? 56 : 47)
+#else
+#define __KASAN_VIRT_SHIFT	(boot_cpu_has(X86_FEATURE_LA57) ? 56 : 47)
+#endif
+
 #define KASAN_SHADOW_START      (KASAN_SHADOW_OFFSET + \
-					((-1UL << __VIRTUAL_MASK_SHIFT) >> \
+					((-1UL << __KASAN_VIRT_SHIFT) >> \
 						KASAN_SHADOW_SCALE_SHIFT))
 /*
  * 47 bits for kernel address -> (47 - KASAN_SHADOW_SCALE_SHIFT) bits for shadow
-- 
  Kiryl Shutsemau / Kirill A. Shutemov