kernel_optimize_test/scripts/link-vmlinux.sh
Stephen Rothwell a5967db9af kbuild: allow architectures to use thin archives instead of ld -r
ld -r is an incremental link used to create built-in.o files in build
subdirectories. It produces relocatable object files containing all
its input files, and these are are then pulled together and relocated
in the final link. Aside from the bloat, this constrains the final
link relocations, which has bitten large powerpc builds with
unresolvable relocations in the final link.

Alan Modra has recommended the kernel use thin archives for linking.
This is an alternative and means that the linker has more information
available to it when it links the kernel.

This patch enables a config option architectures can select, which
causes all built-in.o files to be built as thin archives. built-in.o
files in subdirectories do not get symbol table or index attached,
which improves speed and size. The final link pass creates a
built-in.o archive in the root output directory which includes the
symbol table and index. The linker then uses takes this file to link.

The --whole-archive linker option is required, because the linker now
has visibility to every individual object file, and it will otherwise
just completely avoid including those without external references
(consider a file with EXPORT_SYMBOL or initcall or hardware exceptions
as its only entry points). The traditional built works "by luck" as
built-in.o files are large enough that they're going to get external
references. However this optimisation is unpredictable for the kernel
(due to above external references), ineffective at culling unused, and
costly because the .o files have to be searched for references.
Superior alternatives for link-time culling should be used instead.

Build characteristics for inclink vs thinarc, on a small powerpc64le
pseries VM with a modest .config:

                                  inclink       thinarc
sizes
vmlinux                        15 618 680    15 625 028
sum of all built-in.o          56 091 808     1 054 334
sum excluding root built-in.o                   151 430

find -name built-in.o | xargs rm ; time make vmlinux
real                              22.772s       21.143s
user                              13.280s       13.430s
sys                                4.310s        2.750s

- Final kernel pulled in only about 6K more, which shows how
  ineffective the object file culling is.
- Build performance looks improved due to less pagecache activity.
  On IO constrained systems it could be a bigger win.
- Build size saving is significant.

Side note, the toochain understands archives, so there's some tricks,
$ ar t built-in.o          # list all files you linked with
$ size built-in.o          # and their sizes
$ objdump -d built-in.o    # disassembly (unrelocated) with filenames

Implementation by sfr, minor tweaks by npiggin.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michal Marek <mmarek@suse.com>
2016-09-09 10:31:19 +02:00

303 lines
6.6 KiB
Bash
Executable File

#!/bin/sh
#
# link vmlinux
#
# vmlinux is linked from the objects selected by $(KBUILD_VMLINUX_INIT) and
# $(KBUILD_VMLINUX_MAIN). Most are built-in.o files from top-level directories
# in the kernel tree, others are specified in arch/$(ARCH)/Makefile.
# Ordering when linking is important, and $(KBUILD_VMLINUX_INIT) must be first.
#
# vmlinux
# ^
# |
# +-< $(KBUILD_VMLINUX_INIT)
# | +--< init/version.o + more
# |
# +--< $(KBUILD_VMLINUX_MAIN)
# | +--< drivers/built-in.o mm/built-in.o + more
# |
# +-< ${kallsymso} (see description in KALLSYMS section)
#
# vmlinux version (uname -v) cannot be updated during normal
# descending-into-subdirs phase since we do not yet know if we need to
# update vmlinux.
# Therefore this step is delayed until just before final link of vmlinux.
#
# System.map is generated to document addresses of all kernel symbols
# Error out on error
set -e
# Nice output in kbuild format
# Will be supressed by "make -s"
info()
{
if [ "${quiet}" != "silent_" ]; then
printf " %-7s %s\n" ${1} ${2}
fi
}
# Thin archive build here makes a final archive with
# symbol table and indexes from vmlinux objects, which can be
# used as input to linker.
#
# Traditional incremental style of link does not require this step
#
# built-in.o output file
#
archive_builtin()
{
if [ -n "${CONFIG_THIN_ARCHIVES}" ]; then
info AR built-in.o
rm -f built-in.o;
${AR} rcsT${KBUILD_ARFLAGS} built-in.o \
${KBUILD_VMLINUX_INIT} \
${KBUILD_VMLINUX_MAIN}
fi
}
# Link of vmlinux.o used for section mismatch analysis
# ${1} output file
modpost_link()
{
local objects
if [ -n "${CONFIG_THIN_ARCHIVES}" ]; then
objects="--whole-archive built-in.o"
else
objects="${KBUILD_VMLINUX_INIT} \
--start-group \
${KBUILD_VMLINUX_MAIN} \
--end-group"
fi
${LD} ${LDFLAGS} -r -o ${1} ${objects}
}
# Link of vmlinux
# ${1} - optional extra .o files
# ${2} - output file
vmlinux_link()
{
local lds="${objtree}/${KBUILD_LDS}"
local objects
if [ "${SRCARCH}" != "um" ]; then
if [ -n "${CONFIG_THIN_ARCHIVES}" ]; then
objects="--whole-archive built-in.o ${1}"
else
objects="${KBUILD_VMLINUX_INIT} \
--start-group \
${KBUILD_VMLINUX_MAIN} \
--end-group \
${1}"
fi
${LD} ${LDFLAGS} ${LDFLAGS_vmlinux} -o ${2} \
-T ${lds} ${objects}
else
if [ -n "${CONFIG_THIN_ARCHIVES}" ]; then
objects="-Wl,--whole-archive built-in.o ${1}"
else
objects="${KBUILD_VMLINUX_INIT} \
-Wl,--start-group \
${KBUILD_VMLINUX_MAIN} \
-Wl,--end-group \
${1}"
fi
${CC} ${CFLAGS_vmlinux} -o ${2} \
-Wl,-T,${lds} \
${objects} \
-lutil -lrt -lpthread
rm -f linux
fi
}
# Create ${2} .o file with all symbols from the ${1} object file
kallsyms()
{
info KSYM ${2}
local kallsymopt;
if [ -n "${CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX}" ]; then
kallsymopt="${kallsymopt} --symbol-prefix=_"
fi
if [ -n "${CONFIG_KALLSYMS_ALL}" ]; then
kallsymopt="${kallsymopt} --all-symbols"
fi
if [ -n "${CONFIG_KALLSYMS_ABSOLUTE_PERCPU}" ]; then
kallsymopt="${kallsymopt} --absolute-percpu"
fi
if [ -n "${CONFIG_KALLSYMS_BASE_RELATIVE}" ]; then
kallsymopt="${kallsymopt} --base-relative"
fi
local aflags="${KBUILD_AFLAGS} ${KBUILD_AFLAGS_KERNEL} \
${NOSTDINC_FLAGS} ${LINUXINCLUDE} ${KBUILD_CPPFLAGS}"
local afile="`basename ${2} .o`.S"
${NM} -n ${1} | scripts/kallsyms ${kallsymopt} > ${afile}
${CC} ${aflags} -c -o ${2} ${afile}
}
# Create map file with all symbols from ${1}
# See mksymap for additional details
mksysmap()
{
${CONFIG_SHELL} "${srctree}/scripts/mksysmap" ${1} ${2}
}
sortextable()
{
${objtree}/scripts/sortextable ${1}
}
# Delete output files in case of error
cleanup()
{
rm -f .old_version
rm -f .tmp_System.map
rm -f .tmp_kallsyms*
rm -f .tmp_version
rm -f .tmp_vmlinux*
rm -f built-in.o
rm -f System.map
rm -f vmlinux
rm -f vmlinux.o
}
on_exit()
{
if [ $? -ne 0 ]; then
cleanup
fi
}
trap on_exit EXIT
on_signals()
{
exit 1
}
trap on_signals HUP INT QUIT TERM
#
#
# Use "make V=1" to debug this script
case "${KBUILD_VERBOSE}" in
*1*)
set -x
;;
esac
if [ "$1" = "clean" ]; then
cleanup
exit 0
fi
# We need access to CONFIG_ symbols
case "${KCONFIG_CONFIG}" in
*/*)
. "${KCONFIG_CONFIG}"
;;
*)
# Force using a file from the current directory
. "./${KCONFIG_CONFIG}"
esac
archive_builtin
#link vmlinux.o
info LD vmlinux.o
modpost_link vmlinux.o
# modpost vmlinux.o to check for section mismatches
${MAKE} -f "${srctree}/scripts/Makefile.modpost" vmlinux.o
# Update version
info GEN .version
if [ ! -r .version ]; then
rm -f .version;
echo 1 >.version;
else
mv .version .old_version;
expr 0$(cat .old_version) + 1 >.version;
fi;
# final build of init/
${MAKE} -f "${srctree}/scripts/Makefile.build" obj=init GCC_PLUGINS_CFLAGS="${GCC_PLUGINS_CFLAGS}"
kallsymso=""
kallsyms_vmlinux=""
if [ -n "${CONFIG_KALLSYMS}" ]; then
# kallsyms support
# Generate section listing all symbols and add it into vmlinux
# It's a three step process:
# 1) Link .tmp_vmlinux1 so it has all symbols and sections,
# but __kallsyms is empty.
# Running kallsyms on that gives us .tmp_kallsyms1.o with
# the right size
# 2) Link .tmp_vmlinux2 so it now has a __kallsyms section of
# the right size, but due to the added section, some
# addresses have shifted.
# From here, we generate a correct .tmp_kallsyms2.o
# 2a) We may use an extra pass as this has been necessary to
# woraround some alignment related bugs.
# KALLSYMS_EXTRA_PASS=1 is used to trigger this.
# 3) The correct ${kallsymso} is linked into the final vmlinux.
#
# a) Verify that the System.map from vmlinux matches the map from
# ${kallsymso}.
kallsymso=.tmp_kallsyms2.o
kallsyms_vmlinux=.tmp_vmlinux2
# step 1
vmlinux_link "" .tmp_vmlinux1
kallsyms .tmp_vmlinux1 .tmp_kallsyms1.o
# step 2
vmlinux_link .tmp_kallsyms1.o .tmp_vmlinux2
kallsyms .tmp_vmlinux2 .tmp_kallsyms2.o
# step 2a
if [ -n "${KALLSYMS_EXTRA_PASS}" ]; then
kallsymso=.tmp_kallsyms3.o
kallsyms_vmlinux=.tmp_vmlinux3
vmlinux_link .tmp_kallsyms2.o .tmp_vmlinux3
kallsyms .tmp_vmlinux3 .tmp_kallsyms3.o
fi
fi
info LD vmlinux
vmlinux_link "${kallsymso}" vmlinux
if [ -n "${CONFIG_BUILDTIME_EXTABLE_SORT}" ]; then
info SORTEX vmlinux
sortextable vmlinux
fi
info SYSMAP System.map
mksysmap vmlinux System.map
# step a (see comment above)
if [ -n "${CONFIG_KALLSYMS}" ]; then
mksysmap ${kallsyms_vmlinux} .tmp_System.map
if ! cmp -s System.map .tmp_System.map; then
echo >&2 Inconsistent kallsyms data
echo >&2 Try "make KALLSYMS_EXTRA_PASS=1" as a workaround
exit 1
fi
fi
# We made a new kernel - delete old version file
rm -f .old_version