linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Kees Cook <keescook@chromium.org>
To: Masahiro Yamada <masahiroy@kernel.org>
Cc: linux-kbuild@vger.kernel.org,
	Michal Marek <michal.lkml@markovi.net>,
	Nathan Chancellor <nathan@kernel.org>,
	Nick Desaulniers <ndesaulniers@google.com>,
	clang-built-linux@googlegroups.com, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 08/13] kbuild: build modules in the same way with/without Clang LTO
Date: Tue, 31 Aug 2021 10:39:15 -0700	[thread overview]
Message-ID: <202108311037.B3C7FCAF@keescook> (raw)
In-Reply-To: <20210831074004.3195284-9-masahiroy@kernel.org>

On Tue, Aug 31, 2021 at 04:39:59PM +0900, Masahiro Yamada wrote:
> When Clang LTO is enabled, additional intermediate files *.lto.o are
> created because LLVM bitcode must be converted to ELF before modpost.
> 
> For non-LTO builds:
> 
>          $(LD)             $(LD)
>   objects ---> <modname>.o -----> <modname>.ko
>                              |
>           <modname>.mod.o ---/
> 
> For Clang LTO builds:
> 
>          $(AR)            $(LD)                 $(LD)
>   objects ---> <modname>.o ---> <modname>.lto.o -----> <modname>.ko
>                                                   |
>                                 <modname>.mod.o --/
> 
> Since the Clang LTO introduction, Kbuild code is complicated due to
> CONFIG_LTO_CLANG conditionals sprinkled everywhere.
> 
> Another confusion for Clang LTO builds is, <modname>.o is an archive
> that contains LLVM bitcode files. The suffix should be .a instead of .o
> 
> To clean up the code, unify the build process of modules, as follows:
> 
>          $(AR)            $(LD)                     $(LD)
>   objects ---> <modname>.a ---> <modname>.prelink.o -----> <modname>.ko
>                                                       |
>                                 <modname>.mod.o ------/
> 
> Here, 'objects' are either ELF or LLVM bitcode. <modname>.a is an archive,
> <modname>.prelink.o is ELF.

This is a good diagram and helps me understand what's happening here. Do
you think there's a place for it somewhere in the kbuild documentation?

> 
> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

My question about speed changes also applies to this, since there's now
a new step for non-LTO builds. I think you said it wasn't a meaningful
change in speed, but I think it'd be worth mentioning performance
changes in this commit message.

> ---
> 
>  scripts/Makefile.build    | 100 +++++++++++++++++---------------------
>  scripts/Makefile.lib      |  11 ++---
>  scripts/Makefile.modfinal |   4 +-
>  scripts/Makefile.modpost  |   7 +--
>  scripts/mod/modpost.c     |   6 +--
>  5 files changed, 56 insertions(+), 72 deletions(-)
> 
> diff --git a/scripts/Makefile.build b/scripts/Makefile.build
> index 3ad1b1227371..cdc09e9080ca 100644
> --- a/scripts/Makefile.build
> +++ b/scripts/Makefile.build
> @@ -88,9 +88,7 @@ endif
>  
>  targets-for-modules := $(patsubst %.o, %.mod, $(filter %.o, $(obj-m)))
>  
> -ifdef CONFIG_LTO_CLANG
> -targets-for-modules += $(patsubst %.o, %.lto.o, $(filter %.o, $(obj-m)))
> -endif
> +targets-for-modules += $(patsubst %.o, %.prelink.o, $(filter %.o, $(obj-m)))
>  
>  ifdef need-modorder
>  targets-for-modules += $(obj)/modules.order
> @@ -243,9 +241,12 @@ endif # CONFIG_STACK_VALIDATION
>  
>  ifdef CONFIG_LTO_CLANG
>  
> -# Skip objtool for LLVM bitcode
> +# Skip objtool LLVM bitcode

Nit: needless comment change?

>  $(obj)/%o: objtool-enabled :=
>  
> +# Run objtool now that we have compiled modules into native code
> +$(obj)/%.prelink.o: objtool-enabled := y
> +
>  else
>  
>  # 'OBJECT_FILES_NON_STANDARD := y': skip objtool checking for a directory
> @@ -255,6 +256,8 @@ else
>  $(obj)/%o: objtool-enabled = $(if $(filter-out y%, \
>  	$(OBJECT_FILES_NON_STANDARD_$(basetarget).o)$(OBJECT_FILES_NON_STANDARD)n),y)
>  
> +$(obj)/%.prelink.o: objtool-enabled :=
> +
>  endif
>  
>  ifdef CONFIG_TRIM_UNUSED_KSYMS
> @@ -287,32 +290,12 @@ $(obj)/%.o: $(src)/%.c $(recordmcount_source) FORCE
>  	$(call if_changed_rule,cc_o_c)
>  	$(call cmd,force_checksrc)
>  
> -ifdef CONFIG_LTO_CLANG
> -# Module .o files may contain LLVM bitcode, compile them into native code
> -# before ELF processing
> -quiet_cmd_cc_lto_link_modules = LTO [M] $@
> -cmd_cc_lto_link_modules =						\
> -	$(LD) $(ld_flags) -r -o $@					\
> -		$(shell [ -s $(@:.lto.o=.o.symversions) ] &&		\
> -			echo -T $(@:.lto.o=.o.symversions))		\
> -		--whole-archive $(filter-out FORCE,$^)			\
> -		$(cmd_objtool)
> -
> -# objtool was skipped for LLVM bitcode, run it now that we have compiled
> -# modules into native code
> -$(obj)/%.lto.o: objtool-enabled = y
> -$(obj)/%.lto.o: part-of-module := y
> -
> -$(obj)/%.lto.o: $(obj)/%.o FORCE
> -	$(call if_changed,cc_lto_link_modules)
> -endif
> -
>  cmd_mod = { \
>  	echo $(if $($*-objs)$($*-y)$($*-m), $(addprefix $(obj)/, $($*-objs) $($*-y) $($*-m)), $(@:.mod=.o)); \
>  	$(undefined_syms) echo; \
>  	} > $@
>  
> -$(obj)/%.mod: $(obj)/%$(mod-prelink-ext).o FORCE
> +$(obj)/%.mod: $(obj)/%.prelink.o FORCE
>  	$(call if_changed,mod)
>  
>  quiet_cmd_cc_lst_c = MKLST   $@
> @@ -416,17 +399,6 @@ $(obj)/%.asn1.c $(obj)/%.asn1.h: $(src)/%.asn1 $(objtree)/scripts/asn1_compiler
>  $(subdir-builtin): $(obj)/%/built-in.a: $(obj)/% ;
>  $(subdir-modorder): $(obj)/%/modules.order: $(obj)/% ;
>  
> -# combine symversions for later processing
> -ifeq ($(CONFIG_LTO_CLANG) $(CONFIG_MODVERSIONS),y y)
> -      cmd_update_lto_symversions =					\
> -	rm -f $@.symversions						\
> -	$(foreach n, $(filter-out FORCE,$^),				\
> -		$(if $(shell test -s $(n).symversions && echo y),	\
> -			; cat $(n).symversions >> $@.symversions))
> -else
> -      cmd_update_lto_symversions = echo >/dev/null
> -endif
> -
>  #
>  # Rule to compile a set of .o files into one .a file (without symbol table)
>  #
> @@ -446,10 +418,10 @@ $(obj)/built-in.a: $(real-obj-y) FORCE
>  # modules.order unless contained modules are updated.
>  
>  cmd_modules_order = { $(foreach m, $(real-prereqs), \
> -	$(if $(filter %/modules.order, $m), cat $m, echo $(patsubst %.o,%.ko,$m));) :; } \
> +	$(if $(filter %/modules.order, $m), cat $m, echo $(patsubst %.a,%.ko,$m));) :; } \
>  	| $(AWK) '!x[$$0]++' - > $@
>  
> -$(obj)/modules.order: $(obj-m) FORCE
> +$(obj)/modules.order: $(modules) FORCE
>  	$(call if_changed,modules_order)
>  
>  #
> @@ -458,26 +430,44 @@ $(obj)/modules.order: $(obj-m) FORCE
>  $(obj)/lib.a: $(lib-y) FORCE
>  	$(call if_changed,ar)
>  
> -# NOTE:
> -# Do not replace $(filter %.o,^) with $(real-prereqs). When a single object
> -# module is turned into a multi object module, $^ will contain header file
> -# dependencies recorded in the .*.cmd file.
> -ifdef CONFIG_LTO_CLANG
> -quiet_cmd_link_multi-m = AR [M]  $@
> -cmd_link_multi-m =						\
> -	$(cmd_update_lto_symversions);				\
> -	rm -f $@; 						\
> -	$(AR) cDPrsT $@ $(filter %.o,$^)
> -else
> -quiet_cmd_link_multi-m = LD [M]  $@
> -      cmd_link_multi-m = $(LD) $(ld_flags) -r -o $@ $(filter %.o,$^)
> +#
> +# Rule to prelink modules
> +#
> +
> +ifeq ($(CONFIG_LTO_CLANG) $(CONFIG_MODVERSIONS),y y)
> +
> +cmd_merge_symver = $(PERL) scripts/merge-symvers.pl -a $(AR) -o $@ $<
> +
> +$(obj)/%.prelink.symversions: $(obj)/%.a FORCE
> +	$(call if_changed,merge_symver)
> +
> +targets += $(patsubst %.a, %.prelink.symversions, $(modules))
> +
> +$(obj)/%.prelink.o: ld_flags += --script=$(filter %.symversions,$^)
> +module-symver = $(obj)/%.prelink.symversions
> +
>  endif
>  
> -$(multi-obj-m): FORCE
> -	$(call if_changed,link_multi-m)
> -$(call multi_depend, $(multi-obj-m), .o, -objs -y -m)
> +quiet_cmd_ld_o_a = LD [M]  $@
> +      cmd_ld_o_a = $(LD) $(ld_flags) -r -o $@ --whole-archive $< $(cmd_objtool)
> +
> +$(obj)/%.prelink.o: part-of-module := y
> +
> +$(obj)/%.prelink.o: $(obj)/%.a $(module-symver) FORCE
> +	$(call if_changed,ld_o_a)
> +
> +quiet_cmd_ar_module = AR [M]  $@
> +      cmd_ar_module = rm -f $@; $(AR) cDPrST $@ $(real-prereqs)
> +
> +$(modules-single): %.a: %.o FORCE
> +	$(call if_changed,ar_module)
> +
> +$(modules-multi): FORCE
> +	$(call if_changed,ar_module)
> +$(call multi_depend, $(modules-multi), .a, -objs -y -m)
> +
> +targets += $(modules-single) $(modules-multi)
>  
> -targets += $(multi-obj-m)
>  targets := $(filter-out $(PHONY), $(targets))
>  
>  # Add intermediate targets:
> diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
> index 34c4c11c4bc1..5074922db82d 100644
> --- a/scripts/Makefile.lib
> +++ b/scripts/Makefile.lib
> @@ -106,6 +106,10 @@ multi-dtb-y	:= $(addprefix $(obj)/, $(multi-dtb-y))
>  real-dtb-y	:= $(addprefix $(obj)/, $(real-dtb-y))
>  subdir-ym	:= $(addprefix $(obj)/,$(subdir-ym))
>  
> +modules		:= $(patsubst %.o, %.a, $(obj-m))
> +modules-multi	:= $(sort $(patsubst %.o, %.a, $(multi-obj-m)))
> +modules-single	:= $(sort $(filter-out $(modules-multi), $(filter %.a, $(modules))))
> +
>  # Finds the multi-part object the current object will be linked into.
>  # If the object belongs to two or more multi-part objects, list them all.
>  modname-multi = $(sort $(foreach m,$(multi-obj-ym),\
> @@ -225,13 +229,6 @@ dtc_cpp_flags  = -Wp,-MMD,$(depfile).pre.tmp -nostdinc                    \
>  		 $(addprefix -I,$(DTC_INCLUDE))                          \
>  		 -undef -D__DTS__
>  
> -ifeq ($(CONFIG_LTO_CLANG),y)
> -# With CONFIG_LTO_CLANG, .o files in modules might be LLVM bitcode, so we
> -# need to run LTO to compile them into native code (.lto.o) before further
> -# processing.
> -mod-prelink-ext := .lto
> -endif
> -
>  # Useful for describing the dependency of composite objects
>  # Usage:
>  #   $(call multi_depend, multi_used_targets, suffix_to_remove, suffix_to_add)
> diff --git a/scripts/Makefile.modfinal b/scripts/Makefile.modfinal
> index ff805777431c..1b6401f53662 100644
> --- a/scripts/Makefile.modfinal
> +++ b/scripts/Makefile.modfinal
> @@ -9,7 +9,7 @@ __modfinal:
>  include include/config/auto.conf
>  include $(srctree)/scripts/Kbuild.include
>  
> -# for c_flags and mod-prelink-ext
> +# for c_flags
>  include $(srctree)/scripts/Makefile.lib
>  
>  # find all modules listed in modules.order
> @@ -55,7 +55,7 @@ if_changed_except = $(if $(call newer_prereqs_except,$(2))$(cmd-check),      \
>  
>  
>  # Re-generate module BTFs if either module's .ko or vmlinux changed
> -$(modules): %.ko: %$(mod-prelink-ext).o %.mod.o scripts/module.lds $(if $(KBUILD_BUILTIN),vmlinux) FORCE
> +$(modules): %.ko: %.prelink.o %.mod.o scripts/module.lds $(if $(KBUILD_BUILTIN),vmlinux) FORCE
>  	+$(call if_changed_except,ld_ko_o,vmlinux)
>  ifdef CONFIG_DEBUG_INFO_BTF_MODULES
>  	+$(if $(newer-prereqs),$(call cmd,btf_ko))
> diff --git a/scripts/Makefile.modpost b/scripts/Makefile.modpost
> index eef56d629799..11883b31c615 100644
> --- a/scripts/Makefile.modpost
> +++ b/scripts/Makefile.modpost
> @@ -41,9 +41,6 @@ __modpost:
>  include include/config/auto.conf
>  include $(srctree)/scripts/Kbuild.include
>  
> -# for mod-prelink-ext
> -include $(srctree)/scripts/Makefile.lib
> -
>  MODPOST = scripts/mod/modpost								\
>  	$(if $(CONFIG_MODVERSIONS),-m)							\
>  	$(if $(CONFIG_MODULE_SRCVERSION_ALL),-a)					\
> @@ -128,9 +125,9 @@ endif
>  # Read out modules.order to pass in modpost.
>  # Otherwise, allmodconfig would fail with "Argument list too long".
>  quiet_cmd_modpost = MODPOST $@
> -      cmd_modpost = sed 's/\.ko$$/$(mod-prelink-ext)\.o/' $< | $(MODPOST) -T -
> +      cmd_modpost = sed 's/ko$$/prelink.o/' $< | $(MODPOST) -T -
>  
> -$(output-symdump): $(MODORDER) $(input-symdump) $(modules:.ko=$(mod-prelink-ext).o) FORCE
> +$(output-symdump): $(MODORDER) $(input-symdump) $(modules:ko=prelink.o) FORCE
>  	$(call if_changed,modpost)
>  
>  targets += $(output-symdump)
> diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c
> index a26139aa57fd..56cd9b7a5dd0 100644
> --- a/scripts/mod/modpost.c
> +++ b/scripts/mod/modpost.c
> @@ -2000,9 +2000,9 @@ static void read_symbols(const char *modname)
>  		/* strip trailing .o */
>  		tmp = NOFAIL(strdup(modname));
>  		tmp[strlen(tmp) - 2] = '\0';
> -		/* strip trailing .lto */
> -		if (strends(tmp, ".lto"))
> -			tmp[strlen(tmp) - 4] = '\0';
> +		/* strip trailing .prelink */
> +		if (strends(tmp, ".prelink"))
> +			tmp[strlen(tmp) - 8] = '\0';
>  		mod = new_module(tmp);
>  		free(tmp);
>  	}
> -- 
> 2.30.2
> 

Otherwise, looks good!

-- 
Kees Cook

  reply	other threads:[~2021-08-31 17:39 UTC|newest]

Thread overview: 41+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-31  7:39 [PATCH v2 00/13] kbuild: second round of Clang LTO refactoring Masahiro Yamada
2021-08-31  7:39 ` [PATCH v2 01/13] kbuild: move objtool_args back to scripts/Makefile.build Masahiro Yamada
2021-09-04 18:18   ` Josh Poimboeuf
2021-08-31  7:39 ` [PATCH v2 02/13] kbuild: rename __objtool_obj to objtool Masahiro Yamada
2021-08-31 17:16   ` Nick Desaulniers
2021-08-31 17:29   ` Kees Cook
2021-09-04 18:19   ` Josh Poimboeuf
2021-08-31  7:39 ` [PATCH v2 03/13] kbuild: store the objtool command in *.cmd files Masahiro Yamada
2021-08-31 17:23   ` Nick Desaulniers
2021-08-31 17:30   ` Kees Cook
2021-09-04 18:04   ` Josh Poimboeuf
2021-09-04 18:45     ` Josh Poimboeuf
2021-09-08 16:12       ` Masahiro Yamada
2021-08-31  7:39 ` [PATCH v2 04/13] kbuild: factor out OBJECT_FILES_NON_STANDARD check into a macro Masahiro Yamada
2021-08-31 17:25   ` Nick Desaulniers
2021-08-31 17:31   ` Kees Cook
2021-09-04 18:59   ` Josh Poimboeuf
2021-08-31  7:39 ` [PATCH v2 05/13] kbuild: detect objtool update without using .SECONDEXPANSION Masahiro Yamada
2021-08-31 17:32   ` Kees Cook
2021-09-03  0:43     ` Masahiro Yamada
2021-08-31 17:33   ` Nick Desaulniers
2021-09-02 23:42     ` Masahiro Yamada
2021-09-04 19:04   ` Josh Poimboeuf
2021-08-31  7:39 ` [PATCH v2 06/13] kbuild: reuse $(cmd_objtool) for cmd_cc_lto_link_modules Masahiro Yamada
2021-08-31 17:35   ` Kees Cook
2021-09-03  0:39     ` Masahiro Yamada
2021-09-03  1:49       ` Kees Cook
2021-09-04 19:11   ` Josh Poimboeuf
2021-08-31  7:39 ` [PATCH v2 07/13] kbuild: do not create built-in.a.symversions or lib.a.symversions Masahiro Yamada
2021-08-31 17:36   ` Kees Cook
2021-08-31  7:39 ` [PATCH v2 08/13] kbuild: build modules in the same way with/without Clang LTO Masahiro Yamada
2021-08-31 17:39   ` Kees Cook [this message]
2021-08-31 17:46   ` Nick Desaulniers
2021-08-31  7:40 ` [PATCH v2 09/13] kbuild: add cmd_and_savecmd macro Masahiro Yamada
2021-08-31 17:39   ` Kees Cook
2021-08-31  7:40 ` [PATCH v2 10/13] kbuild: rebuild modules when objtool is updated for CONFIG_LTO_CLANG Masahiro Yamada
2021-08-31 17:40   ` Kees Cook
2021-09-04 19:13   ` Josh Poimboeuf
2021-08-31  7:40 ` [PATCH v2 11/13] kbuild: always postpone CRC links for module versioning Masahiro Yamada
2021-08-31  7:40 ` [PATCH v2 12/13] kbuild: merge cmd_modversions_c and cmd_modversions_S Masahiro Yamada
2021-08-31  7:40 ` [PATCH v2 13/13] kbuild: merge cmd_ar_builtin and cmd_ar_module Masahiro Yamada

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=202108311037.B3C7FCAF@keescook \
    --to=keescook@chromium.org \
    --cc=clang-built-linux@googlegroups.com \
    --cc=linux-kbuild@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=masahiroy@kernel.org \
    --cc=michal.lkml@markovi.net \
    --cc=nathan@kernel.org \
    --cc=ndesaulniers@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).