x86: Allow excluding reset vector code from u-boot

From: Gabe Black <[email protected]>

When running from coreboot we don't want this code.

This version works by ifdef-ing out all of the code that would go into those sections and all the code that refers to it. The sections are then empty, and the linker will either leave them empty for the loader to ignore or remove them entirely.

---
Changes in v4:
- Use CONFIG_SYS_X86_RESET_VECTOR instead CONFIG_NO_RESET_CODE
- Add note about CONFIG_SYS_X86_RESET_VECTOR to README

Changes in v3:
- Fix incorrect repeated line in Makefile

Changes in v2:
- Put CONFIG_NO_RESET_CODE into Makefile instead of source files

 Makefile                   |    8 +++-----
 README                     |    4 ++++
 arch/x86/cpu/Makefile      |    5 +++--
 arch/x86/cpu/u-boot.lds    |    3 +++
 include/configs/coreboot.h |    2 ++
 include/configs/eNET.h     |    3 +++
 6 files changed, 18 insertions(+), 7 deletions(-)

index 1a17be9..dcd3a0c 100644
@@ -230,10 +230,8 @@ endif
 # U-Boot objects....order is important (i.e. start must be first)
 
 OBJS  = $(CPUDIR)/start.o
-ifeq ($(CPU),x86)
-OBJS += $(CPUDIR)/start16.o
-OBJS += $(CPUDIR)/resetvec.o
-endif
+OBJS-$(CONFIG_SYS_X86_RESET_VECTOR) += $(CPUDIR)/start16.o
+OBJS-$(CONFIG_SYS_X86_RESET_VECTOR) += $(CPUDIR)/resetvec.o
 ifeq ($(CPU),ppc4xx)
 OBJS += $(CPUDIR)/resetvec.o
 endif

@@ -241,7 +239,7 @@ ifeq ($(CPU),mpc85xx)
 OBJS += $(CPUDIR)/resetvec.o
 endif
 
-OBJS := $(addprefix $(obj),$(OBJS))
+OBJS := $(addprefix $(obj),$(OBJS) $(OBJS-y))
 
 HAVE_VENDOR_COMMON_LIB = $(if $(wildcard board/$(VENDOR)/common/Makefile),y,n)
 

index 2dc0984..8cb7d07 100644
@@ -3621,6 +3621,10 @@ Low Level (hardware related) configuration options:
 		be used if available. These functions may be faster under some
 		conditions but may increase the binary size.
 
+- CONFIG_SYS_X86_RESET_VECTOR
+		If defined, the x86 reset vector code is included. When
+		running from Coreboot this is not needed.
+
 Freescale QE/FMAN Firmware Support:
 -----------------------------------
 

index 7f1fc18..b0c350c 100644
@@ -28,12 +28,13 @@ include $(TOPDIR)/config.mk
 
 LIB	= $(obj)lib$(CPU).o
 
-START	= start.o start16.o resetvec.o
+START-y	= start.o
+START-$(CONFIG_SYS_X86_RESET_VECTOR) += resetvec.o start16.o
 COBJS	= interrupts.o cpu.o
 
 SRCS	:= $(START:.o=.S) $(SOBJS:.o=.S) $(COBJS:.o=.c)
 OBJS	:= $(addprefix $(obj),$(SOBJS) $(COBJS))
-START	:= $(addprefix $(obj),$(START))
+START	:= $(addprefix $(obj),$(START-y))
 
 all:	$(obj).depend $(START) $(LIB)
 

index a1ecefa..e20bb02 100644
@@ -86,6 +86,8 @@ SECTIONS
 	__bios_start = LOADADDR(.bios);
 	__bios_size = SIZEOF(.bios);
 
+#ifndef CONFIG_NO_RESET_CODE
+
 	/*
 	 * The following expressions place the 16-bit Real-Mode code and
 	 * Reset Vector at the end of the Flash ROM

@@ -95,4 +97,5 @@ SECTIONS
 
 	. = RESET_VEC_LOC;
 	.resetvec : AT (CONFIG_SYS_TEXT_BASE + (CONFIG_SYS_MONITOR_LEN - RESET_SEG_SIZE + RESET_VEC_LOC)) { KEEP(*(.resetvec)); }
+#endif
 }

index cc95e2b..699c65f 100644
@@ -38,6 +38,8 @@
 #undef CONFIG_SHOW_BOOT_PROGRESS
 #define CONFIG_LAST_STAGE_INIT
 
+/* Define this for now, will be removed in a later patch */
+#define CONFIG_SYS_X86_RESET_VECTOR
 
 /*-----------------------------------------------------------------------
  * Watchdog Configuration

index 4b1c219..bf8ba19 100644
@@ -38,6 +38,9 @@
 #define CONFIG_SHOW_BOOT_PROGRESS
 #define CONFIG_LAST_STAGE_INIT
 
+/* Not running from Coreboot, so we need the reset vector code */
+#define CONFIG_SYS_X86_RESET_VECTOR
+
 /*-----------------------------------------------------------------------
  * Watchdog Configuration
  * NOTE: If CONFIG_HW_WATCHDOG is NOT defined, the watchdog jumper on the

1.7.7.3

Now that coreboot doesn't need the start16 code, remove it. We need to remove the CONFIG_SYS_X86_RESET_VECTOR option from coreboot.h also.

---
Changes in v4:
- Remove CONFIG_SYS_X86_RESET_VECTOR from coreboot.h

Changes in v3: None
Changes in v2:
- Add new patch to remove coreboot start16 code.

 board/chromebook-x86/coreboot/coreboot_start16.S |   13 -------------
 include/configs/coreboot.h                       |    3 ---
 2 files changed, 0 insertions(+), 16 deletions(-)

index 9ad06df..6b3d92d 100644
@@ -22,19 +22,6 @@
  * MA 02111-1307 USA
  */
 
-/*
- * 16bit initialization code.
- * This code have to map the area of the boot flash
- * that is used by U-boot to its final destination.
- */
-
-.text
-.section .start16, "ax"
-.code16
-.globl board_init16
-board_init16:
-	jmp	board_init16_ret
-
 .section .bios, "ax"
 .code16
 .globl realmode_reset

index 699c65f..4999477 100644
@@ -38,9 +38,6 @@
 #undef CONFIG_SHOW_BOOT_PROGRESS
 #define CONFIG_LAST_STAGE_INIT
 
-/* Define this for now, will be removed in a later patch */
-#define CONFIG_SYS_X86_RESET_VECTOR
-
 /*-----------------------------------------------------------------------
  * Watchdog Configuration
  */

1.7.7.3

Coreboot boards have an LPC TPM connected, so enable this.

---
Changes in v4:
- Remove CONFIG_NO_RESET_CODE from coreboot.h since this is not needed

Changes in v3: None
Changes in v2: None

 include/configs/coreboot.h |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

index 4999477..362310f 100644
@@ -66,6 +66,10 @@
 					 CONFIG_SYS_SCSI_MAX_LUN)
 #endif
 
+/* Generic TPM interfaced through LPC bus */
+#define CONFIG_GENERIC_LPC_TPM
+#define CONFIG_TPM_TIS_BASE_ADDRESS        0xfed40000
+
 /*-----------------------------------------------------------------------
  * Real Time Clock Configuration
  */

1.7.7.3

Dear Simon Glass,

Dear Simon Glass,

Hm.... as is, this requires that we define CONFIG_SYS_X86_RESET_VECTOR in all x86 board config files, just because a single board configuration does not need this? This makes no sense. Please invert the logic.

Best regards,

Wolfgang Denk

DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"The question of whether a computer can think is no more  interesting
than the question of whether a submarine can swim"
                                                - Edsgar W.  Dijkstra

Hi Wolfgang,

That was how the previous patch worked (CONFIG_NO_RESET_CODE). If required the ifneq in the Makefile though, which you objected to. Or is there another way?

Also Graeme mentioned that all x86 boards will soon use Coreboot, so the option (and the code) may in fact be removed one day. That's why I was happy enough to invert it.

There are actually only 2 boards, enet and coreboot.

Regards,
Simon

Dear Simon Glass,

As usual, there are a number of ways.

Option 1: Define CONFIG_SYS_X86_RESET_VECTOR in a common header file,
so that you can #undef it in a board config file which does not want / need this. This is functionally equivalent with your current patch, but you have to care about thgis settingonly in the files where it is actually interesting.

Option 2: use logig similar to the OBJ-y=... in the Makefile, just
inverted. I know there are examples for this, but I'm too lazu at the moment to search.

Maybe, but the approach is wrong even for a single board.

Best regards,

Wolfgang Denk

DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Committee, n.:  A group of men who individually can do nothing but as
a group decide that nothing can be done.                 - Fred Allen