kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion

When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses
the directory tree to determine which EXPORT_SYMBOL to trim. If an
EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the
second traverse, where some source files are recompiled with their
EXPORT_SYMBOL() tuned into a no-op.

Linus stated negative opinions about this slowness in commits:

 - 5cf0fd591f ("Kbuild: disable TRIM_UNUSED_KSYMS option")
 - a555bdd0c5 ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding")

We can do this better now. The final data structures of EXPORT_SYMBOL
are generated by the modpost stage, so modpost can selectively emit
KSYMTAB entries that are really used by modules.

Commit f73edc8951 ("kbuild: unify two modpost invocations") is another
ground-work to do this in a one-pass algorithm. With the list of modules,
modpost sets sym->used if it is used by a module. modpost emits KSYMTAB
only for symbols with sym->used==true.

BTW, Nicolas explained why the trimming was implemented with recursion:

  https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/

Actually, we never achieved that level of optimization where the chain
reaction of trimming comes into play because:

 - CONFIG_LTO_CLANG cannot remove any unused symbols
 - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux,
   but not modules

If deeper trimming is required, we need to revisit this, but I guess
that is unlikely to happen.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
This commit is contained in:
Masahiro Yamada
2023-06-12 00:50:57 +09:00
parent 700c48b439
commit 5e9e95cc91
11 changed files with 78 additions and 264 deletions

View File

@@ -1193,28 +1193,12 @@ endif
export KBUILD_VMLINUX_LIBS
export KBUILD_LDS := arch/$(SRCARCH)/kernel/vmlinux.lds
# Recurse until adjust_autoksyms.sh is satisfied
PHONY += autoksyms_recursive
ifdef CONFIG_TRIM_UNUSED_KSYMS
# For the kernel to actually contain only the needed exported symbols,
# we have to build modules as well to determine what those symbols are.
# (this can be evaluated only once include/config/auto.conf has been included)
KBUILD_MODULES := 1
autoksyms_recursive: $(build-dir) modules.order
$(Q)$(CONFIG_SHELL) $(srctree)/scripts/adjust_autoksyms.sh \
"$(MAKE) -f $(srctree)/Makefile autoksyms_recursive"
endif
autoksyms_h := $(if $(CONFIG_TRIM_UNUSED_KSYMS), include/generated/autoksyms.h)
quiet_cmd_autoksyms_h = GEN $@
cmd_autoksyms_h = mkdir -p $(dir $@); \
$(CONFIG_SHELL) $(srctree)/scripts/gen_autoksyms.sh $@
$(autoksyms_h):
$(call cmd,autoksyms_h)
# '$(AR) mPi' needs 'T' to workaround the bug of llvm-ar <= 14
quiet_cmd_ar_vmlinux.a = AR $@
cmd_ar_vmlinux.a = \
@@ -1223,7 +1207,7 @@ quiet_cmd_ar_vmlinux.a = AR $@
$(AR) mPiT $$($(AR) t $@ | sed -n 1p) $@ $$($(AR) t $@ | grep -F -f $(srctree)/scripts/head-object-list.txt)
targets += vmlinux.a
vmlinux.a: $(KBUILD_VMLINUX_OBJS) scripts/head-object-list.txt autoksyms_recursive FORCE
vmlinux.a: $(KBUILD_VMLINUX_OBJS) scripts/head-object-list.txt FORCE
$(call if_changed,ar_vmlinux.a)
PHONY += vmlinux_o
@@ -1279,7 +1263,7 @@ scripts: scripts_basic scripts_dtc
PHONY += prepare archprepare
archprepare: outputmakefile archheaders archscripts scripts include/config/kernel.release \
asm-generic $(version_h) $(autoksyms_h) include/generated/utsrelease.h \
asm-generic $(version_h) include/generated/utsrelease.h \
include/generated/compile.h include/generated/autoconf.h remove-stale-files
prepare0: archprepare
@@ -2039,7 +2023,7 @@ clean: $(clean-dirs)
-o -name '*.dtb.S' -o -name '*.dtbo.S' \
-o -name '*.dt.yaml' \
-o -name '*.dwo' -o -name '*.lst' \
-o -name '*.su' -o -name '*.mod' -o -name '*.usyms' \
-o -name '*.su' -o -name '*.mod' \
-o -name '.*.d' -o -name '.*.tmp' -o -name '*.mod.c' \
-o -name '*.lex.c' -o -name '*.tab.[ch]' \
-o -name '*.asn1.[ch]' \