Re: [PATCH v2 1/3] scripts: kconfig: merge_config.sh: refactor from shell/sed/grep to awk

[Nathan Chancellor <nathan@kernel.org>]

Hi Mikko,

Thanks for the update! Just a heads up, please start a new thread for
newer patch revisions, preferring a link back to the earlier revisions
like b4 does. Sending it as a reply to the previous revision makes it
hard to follow in an mbox, as you might be able to see at the bottom of
the lore thread:

https://lore.kernel.org/linux-kbuild/20251229114447.45236-1-mikko.rapeli@linaro.org/

On Wed, Dec 31, 2025 at 10:40:48AM +0200, Mikko Rapeli wrote:
> From: Anders Roxell <anders.roxell@linaro.org>
> 
> merge_config.sh shell/sed/grep loop scales poorly and is slow.
> With Yocto genericarm64 kernel and around 190 config fragments
> the script takes more than 20 minutes to run on a fast build machine.
> Re-implementation with awk does the same job in 10 seconds.
> Using awk since it is likely available in the build environments
> and using perl, python etc would introduce more complex runtime
> dependencies. awk is good enough and lot better than shell/sed/grep.
> 
> Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
> Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org>
> ---
>  scripts/kconfig/merge_config.sh | 168 ++++++++++++++++++++++++--------
>  1 file changed, 128 insertions(+), 40 deletions(-)
> 
> v2: remove unused sed variables, awk from ${AWK} variable,
>     curly brace syntax fix after rebase, triple check that
>     correct revision of patches are used in testing with
>     yocto/bitbake
> 
> v1: https://lore.kernel.org/linux-kbuild/20251229114447.45236-1-mikko.rapeli@linaro.org/T/#t
> 
> diff --git a/scripts/kconfig/merge_config.sh b/scripts/kconfig/merge_config.sh
> index 79c09b378be8..4cefe3cdfc2f 100755
> --- a/scripts/kconfig/merge_config.sh
> +++ b/scripts/kconfig/merge_config.sh
> @@ -16,8 +16,8 @@
>  set -e
>  
>  clean_up() {
> -	rm -f $TMP_FILE
> -	rm -f $MERGE_FILE
> +	rm -f "${TMP_FILE}"
> +	rm -f "${TMP_FILE}.new"
...
> +if [ -z "${AWK}" ]; then

Small nit: This file seems to prefer $VAR over ${VAR} when not using
bash parameter expansion (and that is a little easier for me as a fish
user to read longterm).

> +	# Normalize: strip trailing comments, convert "is not set" to "=n"
> +	function normalize(line) {
> +		if (line == "") return ""
> +		sub(/[[:space:]]+#.*/, "", line)
> +		if (line ~ / is not set$/) {
> +			sub(/^# /, "", line)
> +			sub(/ is not set$/, "=n", line)
> +		}
> +		return line
> +	}

I think this normalization makes it a little harder to read when the
value is changed from "n" to "y".

Prior to this change:

  $ make -j"$(nproc)" ARCH=arm64 CROSS_COMPILE=aarch64-linux- clean defconfig hardening.config
  ...
  Value of CONFIG_SLAB_FREELIST_HARDENED is redefined by fragment kernel/configs/hardening.config:
  Previous value: # CONFIG_SLAB_FREELIST_HARDENED is not set
  New value: CONFIG_SLAB_FREELIST_HARDENED=y
  ...

After this change:

  $ make -j"$(nproc)" ARCH=arm64 CROSS_COMPILE=aarch64-linux- clean defconfig hardening.config
  ...
  Value of CONFIG_SLAB_FREELIST_HARDENED is redefined by fragment kernel/configs/hardening.config:
    Previous value: CONFIG_SLAB_FREELIST_HARDENED=n
    New value:     CONFIG_SLAB_FREELIST_HARDENED=y
  ...

Linus has complained about moving from "is not set" to "n" in the past:

https://lore.kernel.org/CAHk-=wgxcu9DFkXAOAFdDtLWwuv6qb5iV1E69yWE-JEVsd-NFg@mail.gmail.com/

I do like the alignment change for the "new value" line but I think
keeping "is not set" would be a little easier to quickly parse than
"=n".

  Value of CONFIG_SLAB_FREELIST_HARDENED is redefined by fragment kernel/configs/hardening.config:
    Previous value: # CONFIG_SLAB_FREELIST_HARDENED is not set
    New value:     CONFIG_SLAB_FREELIST_HARDENED=y

Might just be a personal preference though.

> +	function warn_builtin(cfg, prev, new) {
> +		if (warnoverride == "true") return
> +		print cfg ": -y passed, will not demote y to m" > "/dev/stderr"
> +		print "  Previous value: " prev > "/dev/stderr"
> +		print "  New value:	 " new > "/dev/stderr"
> +		print "" > "/dev/stderr"
> +	}
> +
> +	function warn_redefined(cfg, prev, new) {
> +		if (warnoverride == "true") return
> +		print "Value of " cfg " is redefined by fragment " mergefile ":" > "/dev/stderr"
> +		print "  Previous value: " prev > "/dev/stderr"
> +		print "  New value:	 " new > "/dev/stderr"
> +		print "" > "/dev/stderr"
> +	}
> +
> +	function warn_redundant(cfg) {
> +		if (warnredun != "true" || warnoverride == "true") return
> +		print "Value of " cfg " is redundant by fragment " mergefile ":" > "/dev/stderr"
> +	}

The use of /dev/stderr seems to introduce a change in behavior when
using 'make -s'.

Prior to this change:

  $ make -sj"$(nproc)" ARCH=arm64 CROSS_COMPILE=aarch64-linux- clean defconfig hardening.config

After this change:

  $ make -sj"$(nproc)" ARCH=arm64 CROSS_COMPILE=aarch64-linux- clean defconfig hardening.config
  Value of CONFIG_COMPAT_VDSO is redefined by fragment kernel/configs/hardening.config:
    Previous value: CONFIG_COMPAT_VDSO=y
    New value:     CONFIG_COMPAT_VDSO=n

  Value of CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT is redefined by fragment kernel/configs/hardening.config:
    Previous value: CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT=n
    New value:     CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT=y
  ...

These should probably just be /dev/stdout?

Cheers,
Nathan