[PATCH v2] kbuild: Support directory targets for building DTBs

Fri Nov 21 11:42:34 PST 2025

On Thu, Nov 20, 2025 at 11:30:33PM -0700, Nathan Chancellor wrote:
> On Thu, Nov 20, 2025 at 02:48:13PM -0600, Rob Herring (Arm) wrote:
> > It is useful to be able to build all the DTBs for a vendor. One can list
> > all the .dts files in a directory and convert those to %.dtb targets,
> > but that doesn't work for base+overlay DTB targets.
> > 
> > Adding the dts subdirectory is straight-forward, but building the
> > DTBs should only happen for certain targets (dtbs, dtbs_check, %.dtb,
> > %.dtbo, and the directory target(s)).
> > 
> > The 'scripts_dtc' rule doesn't really depend on 'dt_binding_schemas',
> > but the directory target only depends on 'scripts' which depends on
> > 'scripts_dtc'.
> > 
> > Signed-off-by: Rob Herring (Arm) <robh at kernel.org>
> > ---
> > Please ack and I'll take this in the DT tree.
> > 
> > I don't really like looking at MAKECMDGOALS, but that's the only way I
> > could come up with that works. Maybe someone knows a better way.
> 
> Yeah, I do worry that just looking for "dtb" in MAKECMDGOALS could
> result in some false positives but in the tree now, I only see one .c
> file that would trigger this logic, so maybe it is not that big of a
> deal?
> 
>   $ fd dtb
>   arch/microblaze/boot/dts/linked_dtb.S
>   arch/mips/include/asm/mach-loongson64/builtin_dtbs.h
>   arch/um/kernel/dtb.c
>   drivers/gpu/drm/ci/dtbs-check.sh
>   scripts/Makefile.dtbinst
>   scripts/Makefile.dtbs
> 
> Unfortunately, my Make-fu is pretty weak right now so I do not have any
> immediate suggestions but Nicolas might... otherwise, we could probably
> ride this for right now and either revert it or forward fix it if
> problems come up in further testing, since this does seem rather useful
> for folks working on device tree files.

well, evaluating MAKECMDGOALS does not look that bad to me.  But I'd
rather like to reduce the use of findstring, see below.

> 
> > v2:
> >  - Convert arm, mips and riscv. The other DT enabled arches don't have 
> >    vendor directories.
> >  - Link to v1: https://lore.kernel.org/all/20251113225952.867138-1-robh@kernel.org/ 
> > 
> > ---
> >  Makefile              | 2 +-
> >  arch/arm/Kbuild       | 2 ++
> >  arch/arm64/Kbuild     | 2 ++
> >  arch/mips/Kbuild      | 2 ++
> >  arch/riscv/Kbuild     | 2 ++
> >  scripts/Makefile.dtbs | 3 +++
> >  6 files changed, 12 insertions(+), 1 deletion(-)
> > 
> > diff --git a/Makefile b/Makefile
> > index 17cfa11ca716..85018d461575 100644
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -1494,7 +1494,7 @@ export CHECK_DTBS=y
> >  endif
> >  
> >  ifneq ($(CHECK_DTBS),)
> > -dtbs_prepare: dt_binding_schemas
> > +scripts_dtc: dt_binding_schemas
> >  endif
> >  
> >  dtbs_check: dtbs
> > diff --git a/arch/arm/Kbuild b/arch/arm/Kbuild
> > index 69de6b6243c7..af7dd53585c3 100644
> > --- a/arch/arm/Kbuild
> > +++ b/arch/arm/Kbuild
> > @@ -10,5 +10,7 @@ obj-y				+= probes/
> >  obj-y				+= net/
> >  obj-y				+= crypto/
> >  
> > +subdir-y += boot/dts
> > +
> >  # for cleaning
> >  subdir- += boot
> > diff --git a/arch/arm64/Kbuild b/arch/arm64/Kbuild
> > index 5bfbf7d79c99..9e9820af48c9 100644
> > --- a/arch/arm64/Kbuild
> > +++ b/arch/arm64/Kbuild
> > @@ -5,5 +5,7 @@ obj-$(CONFIG_XEN)	+= xen/
> >  obj-$(subst m,y,$(CONFIG_HYPERV))	+= hyperv/
> >  obj-$(CONFIG_CRYPTO)	+= crypto/
> >  
> > +subdir-y += boot/dts
> > +
> >  # for cleaning
> >  subdir- += boot
> > diff --git a/arch/mips/Kbuild b/arch/mips/Kbuild
> > index e901bf554483..6125d00cdcef 100644
> > --- a/arch/mips/Kbuild
> > +++ b/arch/mips/Kbuild
> > @@ -21,5 +21,7 @@ ifdef CONFIG_KVM
> >  obj-y += kvm/
> >  endif
> >  
> > +subdir-y += boot/dts
> > +
> >  # for cleaning
> >  subdir- += boot
> > diff --git a/arch/riscv/Kbuild b/arch/riscv/Kbuild
> > index 126fb738fc44..3cf7f84eb287 100644
> > --- a/arch/riscv/Kbuild
> > +++ b/arch/riscv/Kbuild
> > @@ -7,5 +7,7 @@ obj-$(CONFIG_KVM) += kvm/
> >  
> >  obj-$(CONFIG_ARCH_SUPPORTS_KEXEC_PURGATORY) += purgatory/
> >  
> > +subdir-y += boot/dts
> > +
> >  # for cleaning
> >  subdir- += boot
> > diff --git a/scripts/Makefile.dtbs b/scripts/Makefile.dtbs
> > index 2d321b813600..4d0d164df275 100644
> > --- a/scripts/Makefile.dtbs
> > +++ b/scripts/Makefile.dtbs
> > @@ -14,7 +14,10 @@ dtb-y           := $(addprefix $(obj)/, $(dtb-y))
> >  multi-dtb-y     := $(addprefix $(obj)/, $(multi-dtb-y))
> >  real-dtb-y      := $(addprefix $(obj)/, $(real-dtb-y))
> >  
> > +ifneq ($(findstring /dts/,$(MAKECMDGOALS))$(findstring dtb,$(MAKECMDGOALS)),)

Using '$(filter )' instead of '$(findstring )' reduces the risk of false
positives, e.g.:

dtb-targets := %/dts %.dtb %.dtbo dtbs dtbs_check
ifneq ($(findstring /dts/,$(MAKECMDGOALS))$(filter $(dtb-targets),$(MAKECMDGOALS)),)

With (something like) that:
Tested-by: Nicolas Schier <nsc at kernel.org>
Acked-by: Nicolas Schier <nsc at kernel.org>

But [1] seems to break this patch.

[1]: https://lore.kernel.org/linux-kbuild/20251120140645.478623-1-thomas.de_schampheleire@nokia.com

Kind regards,
Nicolas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-riscv/attachments/20251121/bc6d3ba7/attachment.sig>