arm: Have the linker garbage-collect unused functions and variables

This patch activates -ffunction-sections and -fdata-sections for the
compiler and --gc-sections for the linker. This will strip out all
unused functions and static/global variables from the final binaries and
reduce the amount of data we need to read over SPI.

A quick test with ToT images shows a 2.5k (13%) / 10k (29%) / 12k (28%)
reduction on Nyan and 3k (38%) / 23k (50%) / 13k (29%) on Pit,
respectively for bootblock / romstage / ramstage.

Change-Id: I052411d4ad190d0395921ac4d4677341fb91568a
Signed-off-by: Julius Werner <jwerner@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/177111
(cherry picked from commit 5635b138778dea67a5f179e13003132be07f7e59)
Signed-off-by: Isaac Christensen <isaac.christensen@se-eng.com>
Reviewed-on: http://review.coreboot.org/6904
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>

src/arch/arm/Makefile.inc

@@ -62,9 +62,9 @@ bootblock-y += memmove.S
 $(objcbfs)/bootblock.debug: $(src)/arch/arm/bootblock.ld $(obj)/ldoptions $$(bootblock-objs)
 	@printf "    LINK       $(subst $(obj)/,,$(@))\n"
 ifeq ($(CONFIG_COMPILER_LLVM_CLANG),y)
-	$(LD_bootblock) -m armelf_linux_eabi -static -o $@ -L$(obj) $< -T $(src)/arch/arm/bootblock.ld
+	$(LD_bootblock) -m armelf_linux_eabi --gc-sections -static -o $@ -L$(obj) $< -T $(src)/arch/arm/bootblock.ld
 else
-	$(CC_bootblock) $(CFLAGS_bootblock) -nostartfiles -static -o $@ -L$(obj) -T $(src)/arch/arm/bootblock.ld -Wl,--start-group $(bootblock-objs) $(LIBGCC_FILE_NAME_bootblock) -Wl,--end-group
+	$(CC_bootblock) $(CFLAGS_bootblock) -nostartfiles -Wl,--gc-sections -static -o $@ -L$(obj) -T $(src)/arch/arm/bootblock.ld -Wl,--start-group $(bootblock-objs) $(LIBGCC_FILE_NAME_bootblock) -Wl,--end-group
 endif
 
 endif # CONFIG_ARCH_BOOTBLOCK_ARM
@@ -87,9 +87,9 @@ VBOOT_STUB_DEPS += $(obj)/arch/arm/eabi_compat.rmodules_arm.o
 $(objcbfs)/romstage.debug: $$(romstage-objs) $(src)/arch/arm/romstage.ld $(obj)/ldoptions
 	@printf "    LINK       $(subst $(obj)/,,$(@))\n"
 ifeq ($(CONFIG_COMPILER_LLVM_CLANG),y)
-	$(LD_romstage) -nostdlib -nostartfiles -static -o $@ -L$(obj) $(romstage-objs) -T $(src)/arch/arm/romstage.ld
+	$(LD_romstage) -nostdlib -nostartfiles --gc-sections -static -o $@ -L$(obj) $(romstage-objs) -T $(src)/arch/arm/romstage.ld
 else
-	$(CC_romstage) $(CFLAGS_romstage) -nostartfiles -static -o $@ -L$(obj) -T $(src)/arch/arm/romstage.ld -Wl,--start-group $(romstage-objs) $(LIBGCC_FILE_NAME_romstage) -Wl,--end-group
+	$(CC_romstage) $(CFLAGS_romstage) -nostartfiles -Wl,--gc-sections -static -o $@ -L$(obj) -T $(src)/arch/arm/romstage.ld -Wl,--start-group $(romstage-objs) $(LIBGCC_FILE_NAME_romstage) -Wl,--end-group
 endif
 
 endif # CONFIG_ARCH_ROMSTAGE_ARM
@@ -122,9 +122,9 @@ endif
 $(objgenerated)/ramstage.o: $(stages_o) $$(ramstage-objs) $(LIBGCC_FILE_NAME_ramstage)
 	@printf "    CC         $(subst $(obj)/,,$(@))\n"
 ifeq ($(CONFIG_COMPILER_LLVM_CLANG),y)
-	$(LD_ramstage) -m -m armelf_linux_eabi -r -o $@ --wrap __divdi3 --wrap __udivdi3 --wrap __moddi3 --wrap __umoddi3 --wrap __uidiv --start-group $(ramstage-objs) $(LIBGCC_FILE_NAME_ramstage) --end-group
+	$(LD_ramstage) -m -m armelf_linux_eabi --gc-sections -r -o $@ --wrap __divdi3 --wrap __udivdi3 --wrap __moddi3 --wrap __umoddi3 --wrap __uidiv --start-group $(ramstage-objs) $(LIBGCC_FILE_NAME_ramstage) --end-group
 else
-	$(CC_ramstage) $(CFLAGS_ramstage) $(CPPFLAGS_ramstage) -nostdlib -r -o $@ -Wl,--start-group $(stages_o) $(ramstage-objs) $(LIBGCC_FILE_NAME_ramstage) -Wl,--end-group
+	$(CC_ramstage) $(CFLAGS_ramstage) $(CPPFLAGS_ramstage) -nostdlib -Wl,--gc-sections -r -o $@ -Wl,--start-group $(stages_o) $(ramstage-objs) $(LIBGCC_FILE_NAME_ramstage) -Wl,--end-group
 endif
 
 endif # CONFIG_ARCH_RAMSTAGE_ARM

src/arch/arm/bootblock.ld

@@ -28,23 +28,25 @@ PHDRS
 	to_load PT_LOAD;
 }
 
+ENTRY(_start)
 TARGET(binary)
 SECTIONS
 {
-	ROMLOC = CONFIG_BOOTBLOCK_BASE;
+	. = CONFIG_BOOTBLOCK_BASE;
 
-	/* This section might be better named .setup */
+	.bootblock . : {
-	.rom ROMLOC : {
-		_rom = .;
 		*(.start);
-		*(.id);
+		KEEP(*(.id));
 		*(.text);
 		*(.text.*);
-		*(.rom.text);
+		*(.rodata);
-		*(.rom.data);
-		*(.rom.data.*);
 		*(.rodata.*);
-		_erom = .;
+		*(.data);
+		*(.data.*);
+		*(.bss);
+		*(.bss.*);
+		*(.sbss);
+		*(.sbss.*);
 	} : to_load = 0xff
 
 	/DISCARD/ : {
@@ -54,4 +56,4 @@ SECTIONS
 		*(.note.*)
 		*(.ARM.*)
 	}
-}
+}

src/arch/arm/ramstage.ld

@@ -48,7 +48,7 @@ SECTIONS
 	.ctors : {
 		. = ALIGN(0x100);
 		__CTOR_LIST__ = .;
-		*(.ctors);
+		KEEP(*(.ctors));
 		LONG(0);
 		__CTOR_END__ = .;
 	}
@@ -57,13 +57,13 @@ SECTIONS
 		_rodata = .;
 		. = ALIGN(4);
 		pci_drivers = . ;
-		*(.rodata.pci_driver)
+		KEEP(*(.rodata.pci_driver));
 		epci_drivers = . ;
 		cpu_drivers = . ;
-		*(.rodata.cpu_driver)
+		KEEP(*(.rodata.cpu_driver));
 		ecpu_drivers = . ;
 		_bs_init_begin = .;
-		*(.bs_init)
+		KEEP(*(.bs_init));
 		_bs_init_end = .;
 		*(.rodata)
 		*(.rodata.*)
@@ -83,6 +83,7 @@ SECTIONS
 	.data : {
 		_data = .;
 		*(.data)
+		*(.data.*)
 		_edata = .;
 	}
 
@@ -90,26 +91,27 @@ SECTIONS
 	 * initialized on startup. (typically uninitialized global variables)
 	 * crt0.S fills between _bss and _ebss with zeroes.
 	 */
-	_bss = .;
 	.bss . : {
+		_bss = .;
 		*(.bss)
+		*(.bss.*)
 		*(.sbss)
-		*(COMMON)
+		*(.sbss.*)
+		_ebss = .;
 	}
-	_ebss = .;
 	_end = .;
 
 	/* coreboot really "ends" here. Only heap and stack are placed after
 	 * this line.
 	 */
 
-        _heap = .;
         .heap . : {
+			_heap = .;
                 /* Reserve CONFIG_HEAP_SIZE bytes for the heap */
                 . = CONFIG_HEAP_SIZE ;
                 . = ALIGN(4);
+			_eheap = .;
         }
-        _eheap = .;
 
 	/* The ram segment. This includes all memory used by the memory
 	 * resident copy of coreboot, except the tables that are produced on

src/arch/arm/romstage.ld

@@ -38,19 +38,19 @@ SECTIONS
 	. = CONFIG_ROMSTAGE_BASE;
 
 	.romtext . : {
-		_rom = .;
 		_start = .;
 		*(.text.stage_entry.arm);
 		*(.text.startup);
 		*(.text);
+		*(.text.*);
 	} : to_load
 
 	.romdata . : {
 		*(.rodata);
-		*(.machine_param);
+		*(.rodata.*);
 		*(.data);
+		*(.data.*);
 		. = ALIGN(8);
-		_erom = .;
 	}
 
 	/* bss does not contain data, it is just a space that should be zero
@@ -61,12 +61,12 @@ SECTIONS
 		. = ALIGN(8);
 		_bss = .;
 		*(.bss)
+		*(.bss.*)
 		*(.sbss)
-		*(COMMON)
+		*(.sbss.*)
+		_ebss = .;
 	}
 
-	_ebss = .;
-
 	_end = .;
 
 	/* Discard the sections we don't need/want */

toolchain.inc

@@ -57,10 +57,7 @@ ARCHDIR-i386    := x86
 ARCHDIR-x86_32  := x86
 ARCHDIR-arm     := arm
 
-CFLAGS_arm += \
+CFLAGS_arm      := -mno-unaligned-access -ffunction-sections -fdata-sections
-	-mno-unaligned-access\
-	-mthumb\
-	-mthumb-interwork
 
 toolchain_to_dir = \
 	$(foreach arch,$(ARCH_SUPPORTED),\