linux-kernel - Re: [PATCH] kbuild/x86: Use $(KBUILD_AFLAGS) in Kbuild's version of $(as-instr)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date: Thu, 11 Apr 2024 05:34:52 +0100
From: Dmitry Safonov <0x7f454c46@...il.com>
To: Masahiro Yamada <masahiroy@...nel.org>
Cc: Borislav Petkov <bp@...en8.de>, Dave Hansen <dave.hansen@...ux.intel.com>, 
	Ingo Molnar <mingo@...hat.com>, "H. Peter Anvin" <hpa@...or.com>, Thomas Gleixner <tglx@...utronix.de>, x86@...nel.org, 
	linux-kernel@...r.kernel.org, linux-kbuild@...r.kernel.org, 
	Dmitry Safonov <0x7f454c46@...il.com>
Subject: Re: [PATCH] kbuild/x86: Use $(KBUILD_AFLAGS) in Kbuild's version of $(as-instr)

On Thu, 11 Apr 2024 at 00:41, Dmitry Safonov <0x7f454c46@...il.com> wrote:
>
> Please, ignore this version.
>
> the previous one was to add a helper function and that seemed to work,
> but on rework to actually use KBUILD_AFLAGS instead of adding a new
> function I certainly haven't tested it enough.
> Going to prepare version 2, sorry for the noise.

Yeah, in the end I debugged why it doesn't work, for the same machine
as in the patch message:
$ grep KBUILD_AFLAGS include/config/auto.conf.cmd
ifneq "$(KBUILD_AFLAGS)" "-D__ASSEMBLY__ -fno-PIE -m64 "

Notice the trailing space? That's the thing that doesn't match with
the $(KBUILD_AFLAGS) in top-level Makefile.

Adding some more debug:

> --- a/scripts/kconfig/Makefile
> +++ b/scripts/kconfig/Makefile
> @@ -80,8 +80,10 @@ simple-targets := oldconfig allnoconfig allyesconfig allmodconfig \
>         helpnewconfig yes2modconfig mod2yesconfig mod2noconfig
>
>
>  PHONY += $(simple-targets)
> +$(info KBUILD_AFLAGS is "$(KBUILD_AFLAGS)")
>
>  $(simple-targets): $(obj)/conf
> +       $(info KBUILD_AFLAGS for $@ "$(KBUILD_AFLAGS)")
>         $(Q)$< $(silent) --$@ $(Kconfig)
>
>  PHONY += savedefconfig defconfig
> --- a/scripts/kconfig/preprocess.c
> +++ b/scripts/kconfig/preprocess.c
> @@ -79,6 +79,8 @@ static char *env_expand(const char *name)
>         if (!value)
>                 return NULL;
>
> +       fprintf(stderr, "\tvariable '%s' = '%s'\n", name, value);
> +
>         /*
>          * We need to remember all referenced environment variables.
>          * They will be written out to include/config/auto.conf.cmd

Results in
>          variable 'KBUILD_AFLAGS' = '-D__ASSEMBLY__ -fno-PIE -m64 '
(with the trailing space),
> KBUILD_AFLAGS is "-D__ASSEMBLY__ -fno-PIE -m64"
> KBUILD_AFLAGS for syncconfig "-D__ASSEMBLY__ -fno-PIE -m64 "

Ok, so the variable is actually being updated by a common rule in
> scripts/Makefile.lib:KBUILD_AFLAGS += $(subdir-asflags-y)

Which adds a trailing space to $(KBUILD_AFLAGS) and as a result breaks
the comparison in include/config/auto.conf.cmd

So, I see 3 options for the patch in the thread:
1. Simple and ugly: add ifneq to scripts/Makefile.lib and update
KBUILD_AFLAGS only if $(subdir-asflags-y) is not empty.
2. Maybe a better one: remove trailing (and potentially leading, and
duplicate) spaces for expanded variables in
scripts/kconfig/preprocess.c
3. If the issue in the patch description is not worth it: I could just
work that around in the build system.

Please, advise what's your opinion/views.

Thanks,
             Dmitry