From cc03e713d6f0e0cb8f782152ca8478b1bd3225a7 Mon Sep 17 00:00:00 2001
From: Sultan Alsawaf <sultan@kerneltoast.com>
Date: Sat, 26 Dec 2020 19:58:22 -0800
Subject: [PATCH] lto: Add Link Time Optimization support for GCC

This requires a modern version of GCC and various other patches in order
to work. LTO results in a smaller kernel binary with better performance.

Based off of work from Andi Kleen <ak@linux.intel.com>.

Signed-off-by: Sultan Alsawaf <sultan@kerneltoast.com>
---
 Makefile                | 20 ++++++++++++++++++--
 init/Kconfig            |  3 +++
 scripts/Makefile.lib    |  2 +-
 scripts/link-vmlinux.sh |  4 ++--
 4 files changed, 24 insertions(+), 5 deletions(-)

diff --git a/Makefile b/Makefile
index ce823734bea4..5962c09af907 100644
--- a/Makefile
+++ b/Makefile
@@ -380,8 +380,8 @@ LD		= $(CROSS_COMPILE)ld
 CC		= $(CROSS_COMPILE)gcc
 LDGOLD		= $(CROSS_COMPILE)ld.gold
 CPP		= $(CC) -E
-AR		= $(CROSS_COMPILE)ar
-NM		= $(CROSS_COMPILE)nm
+AR             ?= $(CROSS_COMPILE)ar
+NM             ?= $(CROSS_COMPILE)nm
 STRIP		= $(CROSS_COMPILE)strip
 OBJCOPY		= $(CROSS_COMPILE)objcopy
 OBJDUMP		= $(CROSS_COMPILE)objdump
@@ -684,6 +684,22 @@ LLVM_NM		:= llvm-nm
 export LLVM_AR LLVM_NM
 endif
 
+ifdef CONFIG_LTO_GCC
+LTO_CFLAGS	:= -flto -flto=jobserver -fno-fat-lto-objects \
+		   -fuse-linker-plugin -fwhole-program
+KBUILD_CFLAGS	+= $(LTO_CFLAGS)
+LTO_LDFLAGS	:= $(LTO_CFLAGS) -Wno-lto-type-mismatch -Wno-psabi \
+		   -Wno-stringop-overflow -flinker-output=nolto-rel
+LDFINAL		:= $(CONFIG_SHELL) $(srctree)/scripts/gcc-ld $(LTO_LDFLAGS)
+AR		:= $(CROSS_COMPILE)gcc-ar
+NM		:= $(CROSS_COMPILE)gcc-nm
+DISABLE_LTO	:= -fno-lto
+export DISABLE_LTO LDFINAL
+else
+LDFINAL		:= $(LD)
+export LDFINAL
+endif
+
 # The arch Makefile can set ARCH_{CPP,A,C}FLAGS to override the default
 # values of the respective KBUILD_* variables
 ARCH_CPPFLAGS :=
diff --git a/init/Kconfig b/init/Kconfig
index f38d241dc9d5..02487c86acd2 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1209,6 +1209,9 @@ config CC_OPTIMIZE_FOR_SIZE
 
 endchoice
 
+config LTO_GCC
+	bool "Enable Link Time Optimization for GCC"
+
 config SYSCTL
 	bool
 
diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
index 7058964f99f3..e4e1dc0b14ae 100644
--- a/scripts/Makefile.lib
+++ b/scripts/Makefile.lib
@@ -250,7 +250,7 @@ $(obj)/%: $(src)/%_shipped
 # ---------------------------------------------------------------------------
 
 quiet_cmd_ld = LD      $@
-cmd_ld = $(LD) $(LDFLAGS) $(ldflags-y) $(LDFLAGS_$(@F)) \
+cmd_ld = $(LDFINAL) $(LDFLAGS) $(ldflags-y) $(LDFLAGS_$(@F)) \
 	       $(filter-out FORCE,$^) -o $@
 
 # Objcopy
diff --git a/scripts/link-vmlinux.sh b/scripts/link-vmlinux.sh
index c98e0393978c..83afdb0d697a 100644
--- a/scripts/link-vmlinux.sh
+++ b/scripts/link-vmlinux.sh
@@ -125,7 +125,7 @@ modpost_link()
 		info LD vmlinux.o
 	fi
 
-	${LD} ${LDFLAGS} -r -o ${1} $(lto_lds) ${objects}
+	${LDFINAL} ${LDFLAGS} -r -o ${1} $(lto_lds) ${objects}
 }
 
 # If CONFIG_LTO_CLANG is selected, we postpone running recordmcount until
@@ -150,7 +150,7 @@ vmlinux_link()
 	local objects
 
 	if [ "${SRCARCH}" != "um" ]; then
-		local ld=${LD}
+		local ld=${LDFINAL}
 		local ldflags="${LDFLAGS} ${LDFLAGS_vmlinux}"
 
 		if [ -n "${LDFINAL_vmlinux}" ]; then