x86, cpu: Call verify_cpu during 32bit CPU startup
authorKees Cook <kees.cook@canonical.com>
Wed, 10 Nov 2010 18:35:53 +0000 (10:35 -0800)
committerH. Peter Anvin <hpa@linux.intel.com>
Wed, 10 Nov 2010 23:43:09 +0000 (15:43 -0800)
The XD_DISABLE-clearing side-effect needs to happen for both 32bit
and 64bit, but the 32bit init routines were not calling verify_cpu()
yet. This adds that call to gain the side-effect.

The longmode/SSE tests being performed in verify_cpu() need to happen very
early for 64bit but not for 32bit. Instead of including it in two places
for 32bit, we can just include it once in arch/x86/kernel/head_32.S.

Signed-off-by: Kees Cook <kees.cook@canonical.com>
LKML-Reference: <1289414154-7829-4-git-send-email-kees.cook@canonical.com>
Acked-by: Pekka Enberg <penberg@kernel.org>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
arch/x86/kernel/head_32.S
arch/x86/kernel/verify_cpu.S

index bcece91..fdaea52 100644 (file)
@@ -314,6 +314,10 @@ ENTRY(startup_32_smp)
        subl $0x80000001, %eax
        cmpl $(0x8000ffff-0x80000001), %eax
        ja 6f
+
+       /* Clear bogus XD_DISABLE bits */
+       call verify_cpu
+
        mov $0x80000001, %eax
        cpuid
        /* Execute Disable bit supported? */
@@ -609,6 +613,8 @@ ignore_int:
 #endif
        iret
 
+#include "verify_cpu.S"
+
        __REFDATA
 .align 4
 ENTRY(initial_code)
index ccb4136..5644b4b 100644 (file)
@@ -19,6 +19,7 @@
  *
  *     arch/x86_64/boot/compressed/head_64.S: Boot cpu verification
  *     arch/x86_64/kernel/trampoline_64.S: secondary processor verfication
+ *     arch/x86_64/kernel/head_32.S: processor startup
  *
  *     verify_cpu, returns the status of longmode and SSE in register %eax.
  *             0: Success    1: Failure