This patch is purely adding comments. I have verified that the compiled code is bit-for-bit identical. r1059 was a bit overzealous with comment removal.
If anybody complains that this breaks his setup, I need build/arch/x86/intel/core2/stage0.o from a working and from a patched build. If you're seeing differences, your toolchain is broken.
Signed-off-by: Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net
Index: arch/x86/intel/core2/stage0.S =================================================================== --- arch/x86/intel/core2/stage0.S (Revision 1059) +++ arch/x86/intel/core2/stage0.S (Arbeitskopie) @@ -159,9 +159,13 @@ movl $(CACHE_AS_RAM_BASE + CACHE_AS_RAM_SIZE - 4), %eax movl %eax, %esp
- /* Restore the BIST result */ + /* Restore the BIST result. */ movl %ebp, %eax + + /* We need to set ebp? No need. */ movl %esp, %ebp + + /* First parameter: bist */ pushl %eax
#if 0 @@ -170,6 +174,7 @@ #endif
call stage1_phase1 + /* We will not go back. */
port80_post(0x2f) error:
Carl-Daniel Hailfinger wrote:
This patch is purely adding comments. I have verified that the compiled code is bit-for-bit identical. r1059 was a bit overzealous with comment removal.
If anybody complains that this breaks his setup, I need build/arch/x86/intel/core2/stage0.o from a working and from a patched build. If you're seeing differences, your toolchain is broken.
Signed-off-by: Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net
I appreciate your sense for cosmetics. We should really look at setting colons once the whole code is working decently.
Stefan
On 27.11.2008 14:56, Stefan Reinauer wrote:
Carl-Daniel Hailfinger wrote:
This patch is purely adding comments. I have verified that the compiled code is bit-for-bit identical. r1059 was a bit overzealous with comment removal.
If anybody complains that this breaks his setup, I need build/arch/x86/intel/core2/stage0.o from a working and from a patched build. If you're seeing differences, your toolchain is broken.
Signed-off-by: Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net
I appreciate your sense for cosmetics. We should really look at setting colons once the whole code is working decently.
I'm still hoping to get any information about what broke after my partial revert. Since you reverted the comments, I assumed they would be at fault.
My patches fixed genuine bugs in the stage0 and stage1 code and every target except Core2Duo still worked after them. That suggests a problem inside the Core2Duo code. However, without any information about what fails, I can't look for more bugs.
Looking at IRC logs and svn commits, it suddenly seems that r1056 may have broken the code. The changelog suggests that only XIP was enabled, but there also was a CAR area move inside. AFAICS from the IRC timeline, the code worked before r1056 went in.
Regards, Carl-Daniel
-
Carl-Daniel Hailfinger -
Stefan Reinauer