Ronald G. Minnich (rminnich@gmail.com) just uploaded a new patch set to gerrit, which you can find at http://review.coreboot.org/1221
-gerrit
commit 4715d6b4706fb78e9073295b878b69a6cf37ab6f Author: Ronald G. Minnich rminnich@gmail.com Date: Thu Jul 12 11:21:51 2012 -0700
Poison the stack to uncover programming errors
Code can easily make the mistake of using uninitialized values or, in assembly, mistakenly dereferencing stack pointers when an address is desired.
Set the stack to a non-zero value which is also (by testing) a pointer which will crash coreboot if used. This poisoning has uncovered at least one bug.
Change-Id: I4affb9a14b96611e8bf83cb82636e47913025a5d Signed-off-by: Ronald G. Minnich rminnich@gmail.com --- src/arch/x86/lib/c_start.S | 10 +++++++--- 1 files changed, 7 insertions(+), 3 deletions(-)
diff --git a/src/arch/x86/lib/c_start.S b/src/arch/x86/lib/c_start.S index fd9dce7..9c1fdb1 100644 --- a/src/arch/x86/lib/c_start.S +++ b/src/arch/x86/lib/c_start.S @@ -16,17 +16,21 @@ _start:
post_code(POST_ENTRY_C_START) /* post 13 */
- /** clear stack */ + /** poison the stack. Code should not count on the + * stack being full of zeros. This stack poisoning + * recently uncovered a bug in the broadcast SIPI + * code. + */ cld leal _stack, %edi movl $_estack, %ecx subl %edi, %ecx shrl $2, %ecx /* it is 32 bit aligned, right? */ - xorl %eax, %eax + movl $0xDEADBEEF, %eax rep stosl
- /** clear bss */ + /** clear bss, which unlike the stack is zero by definition */ leal _bss, %edi movl $_ebss, %ecx subl %edi, %ecx