[PATCH 0/2] Move tool for generating initramfs cpio to scripts/

[PATCH 0/2] Move tool for generating initramfs cpio to scripts/

[Nicolas Schier]

initramfs cpio tools had been subject of several recent discussions on
linux-kbuild.  Let's get two minor steps forward by enabling a
top-level target for building gen_init_cpio w/o the need of a valid
kbuild configuration [1] and move the tools for generating the builtin
initramfs from usr/ to scripts/ [2].

Link: https://lore.kernel.org/all/20260220191150.244006-1-safinaskar@gmail.com # [1]
Link: https://lore.kernel.org/all/aSdrCFkUQup3qb-q@derry.ads.avm.de/ # [2]
---
Nicolas Schier (2):
      kbuild: Mark usr_gen_init_cpio as no-dot-config-target
      kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

 MAINTAINERS                       | 1 +
 Makefile                          | 9 +++++----
 scripts/.gitignore                | 1 +
 scripts/Makefile                  | 2 ++
 {usr => scripts}/gen_init_cpio.c  | 0
 {usr => scripts}/gen_initramfs.sh | 2 +-
 scripts/remove-stale-files        | 2 ++
 usr/.gitignore                    | 4 +++-
 usr/Makefile                      | 4 +---
 9 files changed, 16 insertions(+), 9 deletions(-)
---
base-commit: 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f
change-id: 20260225-move-gen_init_cpio-to-scripts-9e89ef99e51a

Best regards,
--  
Nicolas

[PATCH 1/2] kbuild: Mark usr_gen_init_cpio as no-dot-config-target

[Nicolas Schier]

Let usr/gen_init_cpio be built also if there is no kernel configuration.

Askar Safin reported that gen_init_cpio may be of use before any kernel
configuration is available.  Add usr_gen_init_cpio to list of targets
that do not require a .config file.

Reported-by: Askar Safin <safinaskar@gmail.com>
Closes: https://lore.kernel.org/all/20260220191150.244006-1-safinaskar@gmail.com
Fixes: f8e05c10631e ("kbuild: Add top-level target for building gen_init_cpio")
Signed-off-by: Nicolas Schier <nsc@kernel.org>
---
 Makefile | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Makefile b/Makefile
index e944c6e71e81..0e788ce45457 100644
--- a/Makefile
+++ b/Makefile
@@ -296,6 +296,7 @@ no-dot-config-targets := $(clean-targets) \
 			 $(version_h) headers headers_% archheaders archscripts \
 			 %asm-generic kernelversion %src-pkg dt_binding_check \
 			 outputmakefile rustavailable rustfmt rustfmtcheck \
+			 usr_gen_init_cpio \
 			 run-command
 no-sync-config-targets := $(no-dot-config-targets) %install modules_sign kernelrelease \
 			  image_name

-- 
2.51.0

[PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Nicolas Schier]

get_init_cpio and gen_initramfs.sh are part of kbuild and required for
all kernel builds.  Move it to scripts/ to be more clear about their
importance.

Link: https://lore.kernel.org/all/aSdrCFkUQup3qb-q@derry.ads.avm.de/
Signed-off-by: Nicolas Schier <nsc@kernel.org>
---

Nathan, could this go through kbuild-fixes even though it is not really
a fix?  Introducing a top-level Makefile target 'usr_gen_init_cpio'
in v7.0 and renaming it to 'scripts/gen_init_cpio` in v7.1 doesn't sound
good to me.
---
 MAINTAINERS                       |  1 +
 Makefile                          | 10 +++++-----
 scripts/.gitignore                |  1 +
 scripts/Makefile                  |  2 ++
 {usr => scripts}/gen_init_cpio.c  |  0
 {usr => scripts}/gen_initramfs.sh |  2 +-
 scripts/remove-stale-files        |  2 ++
 usr/.gitignore                    |  4 +++-
 usr/Makefile                      |  4 +---
 9 files changed, 16 insertions(+), 10 deletions(-)

diff --git a/MAINTAINERS b/MAINTAINERS
index 55af015174a5..0e374c92941f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -13833,6 +13833,7 @@ F:	scripts/basic/
 F:	scripts/clang-tools/
 F:	scripts/container
 F:	scripts/dummy-tools/
+F:	scripts/gen_init*
 F:	scripts/include/
 F:	scripts/install.sh
 F:	scripts/mk*
diff --git a/Makefile b/Makefile
index 0e788ce45457..7798863002ac 100644
--- a/Makefile
+++ b/Makefile
@@ -296,7 +296,7 @@ no-dot-config-targets := $(clean-targets) \
 			 $(version_h) headers headers_% archheaders archscripts \
 			 %asm-generic kernelversion %src-pkg dt_binding_check \
 			 outputmakefile rustavailable rustfmt rustfmtcheck \
-			 usr_gen_init_cpio \
+			 scripts/gen_init_cpio \
 			 run-command
 no-sync-config-targets := $(no-dot-config-targets) %install modules_sign kernelrelease \
 			  image_name
@@ -1437,9 +1437,9 @@ ifdef CONFIG_HEADERS_INSTALL
 prepare: headers
 endif
 
-PHONY += usr_gen_init_cpio
-usr_gen_init_cpio: scripts_basic
-	$(Q)$(MAKE) $(build)=usr usr/gen_init_cpio
+PHONY += scripts/gen_init_cpio
+scripts/gen_init_cpio: scripts_basic
+	$(Q)$(MAKE) $(build)=scripts scripts/gen_init_cpio
 
 PHONY += scripts_unifdef
 scripts_unifdef: scripts_basic
@@ -1709,7 +1709,7 @@ distclean: mrproper
 # Packaging of the kernel to various formats
 # ---------------------------------------------------------------------------
 
-modules-cpio-pkg: usr_gen_init_cpio
+modules-cpio-pkg: scripts/gen_init_cpio
 
 %src-pkg: FORCE
 	$(Q)$(MAKE) -f $(srctree)/scripts/Makefile.package $@
diff --git a/scripts/.gitignore b/scripts/.gitignore
index 4215c2208f7e..700456543910 100644
--- a/scripts/.gitignore
+++ b/scripts/.gitignore
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
 /asn1_compiler
+/gen_init_cpio
 /gen_packed_field_checks
 /generate_rust_target
 /insert-sys-cert
diff --git a/scripts/Makefile b/scripts/Makefile
index 0941e5ce7b57..039863cd064e 100644
--- a/scripts/Makefile
+++ b/scripts/Makefile
@@ -52,6 +52,8 @@ ifdef CONFIG_BUILDTIME_MCOUNT_SORT
 HOSTCFLAGS_sorttable.o += -DMCOUNT_SORT_ENABLED
 endif
 
+hostprogs-always-y += gen_init_cpio
+
 # The following programs are only built on demand
 hostprogs += unifdef gen_packed_field_checks
 
diff --git a/usr/gen_init_cpio.c b/scripts/gen_init_cpio.c
similarity index 100%
rename from usr/gen_init_cpio.c
rename to scripts/gen_init_cpio.c
diff --git a/usr/gen_initramfs.sh b/scripts/gen_initramfs.sh
similarity index 99%
rename from usr/gen_initramfs.sh
rename to scripts/gen_initramfs.sh
index 7eba2fddf0ef..27ac1fd48001 100755
--- a/usr/gen_initramfs.sh
+++ b/scripts/gen_initramfs.sh
@@ -247,4 +247,4 @@ done
 
 # If output_file is set we will generate cpio archive
 # we are careful to delete tmp files
-usr/gen_init_cpio $output $timestamp $cpio_list
+scripts/gen_init_cpio $output $timestamp $cpio_list
diff --git a/scripts/remove-stale-files b/scripts/remove-stale-files
index 6e39fa8540df..e46337e4bfb8 100755
--- a/scripts/remove-stale-files
+++ b/scripts/remove-stale-files
@@ -26,3 +26,5 @@ rm -f scripts/selinux/genheaders/genheaders
 rm -f *.spec
 
 rm -f lib/test_fortify.log
+
+rm -f usr/gen_init_cpio
diff --git a/usr/.gitignore b/usr/.gitignore
index 8996e7a88902..13bf7a38355c 100644
--- a/usr/.gitignore
+++ b/usr/.gitignore
@@ -1,4 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
-/gen_init_cpio
 /initramfs_data.cpio
 /initramfs_inc_data
+
+# Keep around until cleanup in scripts/remove-stale-files
+/gen_init_cpio
diff --git a/usr/Makefile b/usr/Makefile
index e8f42478a0b7..5b97ee66c00a 100644
--- a/usr/Makefile
+++ b/usr/Makefile
@@ -16,8 +16,6 @@ obj-$(CONFIG_BLK_DEV_INITRD) := initramfs_data.o
 
 $(obj)/initramfs_data.o: $(obj)/initramfs_inc_data
 
-hostprogs := gen_init_cpio
-
 ramfs-input := $(CONFIG_INITRAMFS_SOURCE)
 cpio-data :=
 
@@ -72,7 +70,7 @@ quiet_cmd_initfs = GEN     $@
 # 2) There are changes in which files are included (added or deleted)
 # 3) If gen_init_cpio are newer than initramfs_data.cpio
 # 4) Arguments to gen_initramfs.sh changes
-$(obj)/initramfs_data.cpio: $(src)/gen_initramfs.sh $(obj)/gen_init_cpio $(deps_initramfs) FORCE
+$(obj)/initramfs_data.cpio: $(srctree)/scripts/gen_initramfs.sh scripts/gen_init_cpio $(deps_initramfs) FORCE
 	$(call if_changed,initfs)
 
 endif

-- 
2.51.0

Re: [PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Thomas Weißschuh]

On Mon, Mar 09, 2026 at 08:56:30AM +0100, Nicolas Schier wrote:
> get_init_cpio and gen_initramfs.sh are part of kbuild and required for

"gen_init_cpio"

> all kernel builds.  Move it to scripts/ to be more clear about their
> importance.

As far as I can see, gen_init_cpio is currently only built and used if
CONFIG_BLK_DEV_INITRD is enabled. So "required for all kernel builds"
sounds wrong.

This on-demand build can probably be preserved like this:

hostprogs += gen_init_cpio
always-$(CONFIG_BLK_DEV_INITRD) += gen_init_cpio

> Link: https://lore.kernel.org/all/aSdrCFkUQup3qb-q@derry.ads.avm.de/
> Signed-off-by: Nicolas Schier <nsc@kernel.org>
> ---
> 
> Nathan, could this go through kbuild-fixes even though it is not really
> a fix?  Introducing a top-level Makefile target 'usr_gen_init_cpio'
> in v7.0 and renaming it to 'scripts/gen_init_cpio` in v7.1 doesn't sound
> good to me.
> ---
>  MAINTAINERS                       |  1 +
>  Makefile                          | 10 +++++-----
>  scripts/.gitignore                |  1 +
>  scripts/Makefile                  |  2 ++
>  {usr => scripts}/gen_init_cpio.c  |  0
>  {usr => scripts}/gen_initramfs.sh |  2 +-
>  scripts/remove-stale-files        |  2 ++
>  usr/.gitignore                    |  4 +++-
>  usr/Makefile                      |  4 +---
>  9 files changed, 16 insertions(+), 10 deletions(-)
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 55af015174a5..0e374c92941f 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -13833,6 +13833,7 @@ F:	scripts/basic/
>  F:	scripts/clang-tools/
>  F:	scripts/container
>  F:	scripts/dummy-tools/
> +F:	scripts/gen_init*
>  F:	scripts/include/
>  F:	scripts/install.sh
>  F:	scripts/mk*
> diff --git a/Makefile b/Makefile
> index 0e788ce45457..7798863002ac 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -296,7 +296,7 @@ no-dot-config-targets := $(clean-targets) \
>  			 $(version_h) headers headers_% archheaders archscripts \
>  			 %asm-generic kernelversion %src-pkg dt_binding_check \
>  			 outputmakefile rustavailable rustfmt rustfmtcheck \
> -			 usr_gen_init_cpio \
> +			 scripts/gen_init_cpio \
>  			 run-command
>  no-sync-config-targets := $(no-dot-config-targets) %install modules_sign kernelrelease \
>  			  image_name
> @@ -1437,9 +1437,9 @@ ifdef CONFIG_HEADERS_INSTALL
>  prepare: headers
>  endif
>  
> -PHONY += usr_gen_init_cpio
> -usr_gen_init_cpio: scripts_basic
> -	$(Q)$(MAKE) $(build)=usr usr/gen_init_cpio
> +PHONY += scripts/gen_init_cpio
> +scripts/gen_init_cpio: scripts_basic
> +	$(Q)$(MAKE) $(build)=scripts scripts/gen_init_cpio

The other proxy targets for scripts/ use the 'scripts_' prefix.
Why is this one different?

(...)

> diff --git a/scripts/Makefile b/scripts/Makefile
> index 0941e5ce7b57..039863cd064e 100644
> --- a/scripts/Makefile
> +++ b/scripts/Makefile
> @@ -52,6 +52,8 @@ ifdef CONFIG_BUILDTIME_MCOUNT_SORT
>  HOSTCFLAGS_sorttable.o += -DMCOUNT_SORT_ENABLED
>  endif
>  
> +hostprogs-always-y += gen_init_cpio
> +
>  # The following programs are only built on demand
>  hostprogs += unifdef gen_packed_field_checks

(...)

Re: [PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Nicolas Schier]

On Mon, Mar 09, 2026 at 02:41:53PM +0100, Thomas Weißschuh wrote:
> On Mon, Mar 09, 2026 at 08:56:30AM +0100, Nicolas Schier wrote:
> > get_init_cpio and gen_initramfs.sh are part of kbuild and required for
> 
> "gen_init_cpio"

thanks!

> > all kernel builds.  Move it to scripts/ to be more clear about their
> > importance.
> 
> As far as I can see, gen_init_cpio is currently only built and used if
> CONFIG_BLK_DEV_INITRD is enabled. So "required for all kernel builds"
> sounds wrong.
> 
> This on-demand build can probably be preserved like this:
> 
> hostprogs += gen_init_cpio
> always-$(CONFIG_BLK_DEV_INITRD) += gen_init_cpio

Ack, thanks!  I'll take your suggestion.

...
> > diff --git a/Makefile b/Makefile
> > index 0e788ce45457..7798863002ac 100644
> > --- a/Makefile
> > +++ b/Makefile
...
> > @@ -1437,9 +1437,9 @@ ifdef CONFIG_HEADERS_INSTALL
> >  prepare: headers
> >  endif
> >  
> > -PHONY += usr_gen_init_cpio
> > -usr_gen_init_cpio: scripts_basic
> > -	$(Q)$(MAKE) $(build)=usr usr/gen_init_cpio
> > +PHONY += scripts/gen_init_cpio
> > +scripts/gen_init_cpio: scripts_basic
> > +	$(Q)$(MAKE) $(build)=scripts scripts/gen_init_cpio
> 
> The other proxy targets for scripts/ use the 'scripts_' prefix.
> Why is this one different?

Good question; I was ambiguous about that, but I chose 'scripts/' in
favor of 'scripts_' as

1. The $(obj)/initramfs_data.cpio target in usr/Makefile has to depend
   on 'scripts/gen_init_cpio' regardless of how the top-level target is
   called, as the sub-make cannot depend on a phony top-level rule.
   If possible, I like to have the same name for dependencies as for the
   rule that generates them.

2. I cannot see the reason for the 'scripts_' prefix currently used in
   top-level Makefile for real targets building a single output file.
   It seems to me that commit bdd7714b6f4c ("kbuild: build all
   prerequisites of headers_install simultaneously") introduced the
   'scripts_' prefix due to the historic origin 'scripts_basic'; which
   itself came with commit 952a0ae394f4 ("[PATCH] Fix early parallel
   make failures", v2.6.5) [1].  I _guess_ the 'scripts_' prefix was
   meant to clearly make a difference from the previous 'scripts'
   target, and the name makes clear that it is a phony target building
   several things.

Masahiro, if you are around I'd appreciate a comment from you.

[1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=952a0ae394f4

Thomas, does the reasoning make sense to you?  Or do you expect troubles
due to the 'scripts/' prefix?


Thanks for the feedback!

Kind regards,
Nicolas

Re: [PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Thomas Weißschuh]

On Tue, Mar 10, 2026 at 08:14:36AM +0100, Nicolas Schier wrote:
> On Mon, Mar 09, 2026 at 02:41:53PM +0100, Thomas Weißschuh wrote:
> > On Mon, Mar 09, 2026 at 08:56:30AM +0100, Nicolas Schier wrote:

(...)

> > > @@ -1437,9 +1437,9 @@ ifdef CONFIG_HEADERS_INSTALL
> > >  prepare: headers
> > >  endif
> > >  
> > > -PHONY += usr_gen_init_cpio
> > > -usr_gen_init_cpio: scripts_basic
> > > -	$(Q)$(MAKE) $(build)=usr usr/gen_init_cpio
> > > +PHONY += scripts/gen_init_cpio
> > > +scripts/gen_init_cpio: scripts_basic
> > > +	$(Q)$(MAKE) $(build)=scripts scripts/gen_init_cpio
> > 
> > The other proxy targets for scripts/ use the 'scripts_' prefix.
> > Why is this one different?
> 
> Good question; I was ambiguous about that, but I chose 'scripts/' in
> favor of 'scripts_' as
> 
> 1. The $(obj)/initramfs_data.cpio target in usr/Makefile has to depend
>    on 'scripts/gen_init_cpio' regardless of how the top-level target is
>    called, as the sub-make cannot depend on a phony top-level rule.
>    If possible, I like to have the same name for dependencies as for the
>    rule that generates them.
> 
> 2. I cannot see the reason for the 'scripts_' prefix currently used in
>    top-level Makefile for real targets building a single output file.
>    It seems to me that commit bdd7714b6f4c ("kbuild: build all
>    prerequisites of headers_install simultaneously") introduced the
>    'scripts_' prefix due to the historic origin 'scripts_basic'; which
>    itself came with commit 952a0ae394f4 ("[PATCH] Fix early parallel
>    make failures", v2.6.5) [1].  I _guess_ the 'scripts_' prefix was
>    meant to clearly make a difference from the previous 'scripts'
>    target, and the name makes clear that it is a phony target building
>    several things.
> 
> Masahiro, if you are around I'd appreciate a comment from you.
> 
> [1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=952a0ae394f4
> 
> Thomas, does the reasoning make sense to you?  Or do you expect troubles
> due to the 'scripts/' prefix?

Sounds reasonable. Personally I would still prefer consistency, but that is
subjective of course.

In any case:
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>


Thomas

Re: [PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Nicolas Schier]

On Tue, Mar 10, 2026 at 10:13:29AM +0100, Thomas Weißschuh wrote:
> On Tue, Mar 10, 2026 at 08:14:36AM +0100, Nicolas Schier wrote:
> > On Mon, Mar 09, 2026 at 02:41:53PM +0100, Thomas Weißschuh wrote:
> > > On Mon, Mar 09, 2026 at 08:56:30AM +0100, Nicolas Schier wrote:
> 
> (...)
> 
> > > > @@ -1437,9 +1437,9 @@ ifdef CONFIG_HEADERS_INSTALL
> > > >  prepare: headers
> > > >  endif
> > > >  
> > > > -PHONY += usr_gen_init_cpio
> > > > -usr_gen_init_cpio: scripts_basic
> > > > -	$(Q)$(MAKE) $(build)=usr usr/gen_init_cpio
> > > > +PHONY += scripts/gen_init_cpio
> > > > +scripts/gen_init_cpio: scripts_basic
> > > > +	$(Q)$(MAKE) $(build)=scripts scripts/gen_init_cpio
> > > 
> > > The other proxy targets for scripts/ use the 'scripts_' prefix.
> > > Why is this one different?
> > 
> > Good question; I was ambiguous about that, but I chose 'scripts/' in
> > favor of 'scripts_' as
> > 
> > 1. The $(obj)/initramfs_data.cpio target in usr/Makefile has to depend
> >    on 'scripts/gen_init_cpio' regardless of how the top-level target is
> >    called, as the sub-make cannot depend on a phony top-level rule.
> >    If possible, I like to have the same name for dependencies as for the
> >    rule that generates them.
> > 
> > 2. I cannot see the reason for the 'scripts_' prefix currently used in
> >    top-level Makefile for real targets building a single output file.
> >    It seems to me that commit bdd7714b6f4c ("kbuild: build all
> >    prerequisites of headers_install simultaneously") introduced the
> >    'scripts_' prefix due to the historic origin 'scripts_basic'; which
> >    itself came with commit 952a0ae394f4 ("[PATCH] Fix early parallel
> >    make failures", v2.6.5) [1].  I _guess_ the 'scripts_' prefix was
> >    meant to clearly make a difference from the previous 'scripts'
> >    target, and the name makes clear that it is a phony target building
> >    several things.
> > 
> > Masahiro, if you are around I'd appreciate a comment from you.
> > 
> > [1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=952a0ae394f4
> > 
> > Thomas, does the reasoning make sense to you?  Or do you expect troubles
> > due to the 'scripts/' prefix?
> 
> Sounds reasonable. Personally I would still prefer consistency, but that is
> subjective of course.

I'll be considering that.

> In any case:
> Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>

Thanks!

Re: [PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Askar Safin]

Nicolas Schier <nsc@kernel.org>:
> +# Keep around until cleanup in scripts/remove-stale-files

Linus said here https://lore.kernel.org/all/CAHk-=wiHK5_oBUdUiNAaevmN9f-ORe+QBqbRefAZaw-RbgEn3w@mail.gmail.com/
that he doesn't like scripts/remove-stale-files (this is my interpretation).

-- 
Askar Safin

Re: [PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Nicolas Schier]

On Tue, Mar 10, 2026 at 04:49:12AM +0300, Askar Safin wrote:
> Nicolas Schier <nsc@kernel.org>:
> > +# Keep around until cleanup in scripts/remove-stale-files
> 
> Linus said here https://lore.kernel.org/all/CAHk-=wiHK5_oBUdUiNAaevmN9f-ORe+QBqbRefAZaw-RbgEn3w@mail.gmail.com/
> that he doesn't like scripts/remove-stale-files (this is my interpretation).

thanks for the pointer!

Yes, remove-stale-files is suboptimal, but not adding entries there
doesn't make anything better.  And yes, entries in remove-stale-files
should have a life-time of years rather than months; and removing
entries from .gitignore before the remove-stale-files cleanup should be
prevented.

Therefore I chose to add the comment to usr/.gitignore to prevent
accidental removal of the '/gen_init_cpio' before the time has come.

Does that match for you?

Kind regards,
Nicolas

Re: [PATCH 2/2] kbuild: Move gen_init_cpio and gen_initramfs.sh to scripts/

[Askar Safin]

On Tue, Mar 10, 2026 at 10:15 AM Nicolas Schier <nsc@kernel.org> wrote:
> Therefore I chose to add the comment to usr/.gitignore to prevent
> accidental removal of the '/gen_init_cpio' before the time has come.
>
> Does that match for you?

Yes


-- 
Askar Safin

Re: [PATCH 0/2] Move tool for generating initramfs cpio to scripts/

[Askar Safin]

I like this patchset.

Rob (CC'd), does this solve your problem?

Also, in [1] I said:
> command
> "make usr/gen_init_cpio" doesn't work in clean kernel tree even if
> config exists (I checked this)

Now I tested this again, and I see that "make usr/gen_init_cpio" indeed
doesn't work in current mainline in that scenario, but
"cd usr; make gen_init_cpio" works (how I supposed to know this?).
So, problems in [1] are not as big as I thought. You don't need to build
whole kernel two times in that scenario. But still I like this your
new patchset.

Also, I think that top-level "usr" directory exists as a placeholder for
klibc-init project (this is merely my hypothesis, I don't know for sure).
I. e. it was supposed to contain source for userspace init program, which
was supposed to be put into default initramfs.

But klibc-init project, unfortunately, seems to fail. So, I think we should
remove whole top level "usr" directory and move its contents to other places.

[1] https://lore.kernel.org/all/20260220191150.244006-1-safinaskar@gmail.com



-- 
Askar Safin

Re: [PATCH 0/2] Move tool for generating initramfs cpio to scripts/

[H. Peter Anvin]

On March 9, 2026 6:54:16 PM PDT, Askar Safin <safinaskar@gmail.com> wrote:
>I like this patchset.
>
>Rob (CC'd), does this solve your problem?
>
>Also, in [1] I said:
>> command
>> "make usr/gen_init_cpio" doesn't work in clean kernel tree even if
>> config exists (I checked this)
>
>Now I tested this again, and I see that "make usr/gen_init_cpio" indeed
>doesn't work in current mainline in that scenario, but
>"cd usr; make gen_init_cpio" works (how I supposed to know this?).
>So, problems in [1] are not as big as I thought. You don't need to build
>whole kernel two times in that scenario. But still I like this your
>new patchset.
>
>Also, I think that top-level "usr" directory exists as a placeholder for
>klibc-init project (this is merely my hypothesis, I don't know for sure).
>I. e. it was supposed to contain source for userspace init program, which
>was supposed to be put into default initramfs.
>
>But klibc-init project, unfortunately, seems to fail. So, I think we should
>remove whole top level "usr" directory and move its contents to other places.
>
>[1] https://lore.kernel.org/all/20260220191150.244006-1-safinaskar@gmail.com
>
>
>

Well, it worked, but got vetoed. It is still maintained out of tree.