[patch] small fix to espfix code

[patch] small fix to espfix code

[Stas Sergeev]

[-- Attachment #1: Type: text/plain, Size: 846 bytes --]

Hi.

The attached patch fixes a tiny
bug introduced by this patch:
http://lkml.org/lkml/2006/9/30/134
The patch was changing the espfix
stack to be 32bit instead of 16bit,
but this place seems to have been
overlooked.
The code was like this:
---
  	pushl %esp
  	movzwl %sp, %esp
  	addw $4, (%esp)
---
and the patch did:
---
  	pushl %esp
  	CFI_ADJUST_CFA_OFFSET 4
-	movzwl %sp, %esp
  	addw $4, (%esp)
---
but the addw was forgotten to adjust.

The bug is mostly theoretical, I can't
really test the effect of this patch.
So, it is completely untested.

---
Fix tiny bug in the espfix NMI handling
code. The bug was inroduced by this patch:
http://lkml.org/lkml/2006/9/30/134
and can practically never be triggered.

Signed-off-by: Stas Sergeev <stsp@aknet.ru>
CC: Zachary Amsden <zach@vmware.com>
CC: Chuck Ebbert <76306.1226@compuserve.com>

[-- Attachment #2: espfix.diff --]
[-- Type: text/x-patch, Size: 326 bytes --]

--- a/arch/x86/kernel/entry_32.S	2009-01-10 21:03:16.000000000 +0300
+++ b/arch/x86/kernel/entry_32.S	2009-02-15 22:34:23.000000000 +0300
@@ -881,7 +881,7 @@
 	CFI_ADJUST_CFA_OFFSET 4
 	pushl %esp
 	CFI_ADJUST_CFA_OFFSET 4
-	addw $4, (%esp)
+	addl $4, (%esp)
 	/* copy the iret frame of 12 bytes */
 	.rept 3
 	pushl 16(%esp)

Re: [patch] small fix to espfix code

[H. Peter Anvin]

Stas Sergeev wrote:
> 
> The bug is mostly theoretical, I can't
> really test the effect of this patch.
> So, it is completely untested.
> 

It's not a bug, as the stack is aligned such that 16-bit overflow can't 
happen.  However, it is arguably cleaner to have addl here.

	-hpa