Skip to contentSkip to conversation

dts: add property removal option CONFIG_OF_REMOVE_PROPS

PATCH Superseded Anatolij Gustschin 4 participants 10 messages Jan 8, 2020 – Jan 12, 2020 Kbuild

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

This can be used for device tree size reduction similar as CONFIG_OF_SPL_REMOVE_PROPS option. For tbs2910 board this shrinks the image size: all -2304.0 bss -16.0 text -2288.0

---
 dts/Kconfig          | 8 ++++++++
 scripts/Makefile.lib | 3 ++-
 2 files changed, 10 insertions(+), 1 deletion(-)
dts/Kconfig +8 -0
index 64c98dd723..49224bee93 100644
@@ -311,6 +311,14 @@ config OF_SPL_REMOVE_PROPS
 	  can be discarded. This option defines the list of properties to
 	  discard.
 
+config OF_REMOVE_PROPS
+	string "List of device tree properties to drop"
+	depends on OF_CONTROL
+	default "interrupt-parent interrupts" if PINCTRL
+	help
+	  Some properties are not used by U-Boot and can be discarded.
+	  This option defines the list of properties to discard.
+
 config SPL_OF_PLATDATA
 	bool "Generate platform data for use in SPL"
 	depends on SPL_OF_CONTROL
scripts/Makefile.lib +2 -1
index 4ea898a421..3b124369ad 100644
@@ -235,7 +235,8 @@ $(obj)/%.tab.h: $(src)/%.y FORCE
 # ===========================================================================
 
 quiet_cmd_shipped = SHIPPED $@
-cmd_shipped = cat $< > $@
+cmd_shipped = cat $< | $(objtree)/tools/fdtgrep -r -O dtb - -o $@	\
+		$(addprefix -P ,$(subst $\",,$(CONFIG_OF_REMOVE_PROPS)))
 
 $(obj)/%: $(src)/%_shipped
 	$(call cmd,shipped)
Signature · 1 line
2.17.1
#1 lore raw

PATCH 2/2 tbs2910: add custom CONFIG_OF_REMOVE_PROPS list to defconfig

This shrinks the image size: all -1536.0 bss +16.0 text -1552.0

---
NOTE: building with this patch only works when patch [1] is applied.

[1] http://patchwork.ozlabs.org/patch/1219944

 configs/tbs2910_defconfig | 1 +
 1 file changed, 1 insertion(+)
configs/tbs2910_defconfig +1 -0
index a42f53e849..12f3a5dc24 100644
@@ -46,6 +46,7 @@ CONFIG_EFI_PARTITION=y
 CONFIG_OF_CONTROL=y
 CONFIG_OF_EMBED=y
 CONFIG_DEFAULT_DEVICE_TREE="imx6q-tbs2910"
+CONFIG_OF_REMOVE_PROPS="dmas dma-names interrupt-parent interrupts interrupts-extended interrupt-names interrupt-map interrupt-map-mask"
 CONFIG_ENV_IS_IN_MMC=y
 CONFIG_SYS_RELOC_GD_ENV_ADDR=y
 CONFIG_DM=y
Signature · 1 line
2.17.1
#3 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

In reply to Anatolij Gustschin
Quoted 5 lines · Anatolij Gustschin

This can be used for device tree size reduction similar as CONFIG_OF_SPL_REMOVE_PROPS option. For tbs2910 board this shrinks the image size: all -2304.0 bss -16.0 text -2288.0

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200108/32860473/attachment.sig>
#4 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

In reply to Anatolij Gustschin
Quoted 34 lines · Anatolij Gustschin

This can be used for device tree size reduction similar as CONFIG_OF_SPL_REMOVE_PROPS option. For tbs2910 board this shrinks the image size: all -2304.0 bss -16.0 text -2288.0

---
 dts/Kconfig          | 8 ++++++++
 scripts/Makefile.lib | 3 ++-
 2 files changed, 10 insertions(+), 1 deletion(-)
dts/Kconfig +8 -0
index 64c98dd723..49224bee93 100644
@@ -311,6 +311,14 @@ config OF_SPL_REMOVE_PROPS
 	  can be discarded. This option defines the list of properties to
 	  discard.
 
+config OF_REMOVE_PROPS
+	string "List of device tree properties to drop"
+	depends on OF_CONTROL
+	default "interrupt-parent interrupts" if PINCTRL
+	help
+	  Some properties are not used by U-Boot and can be discarded.
+	  This option defines the list of properties to discard.
+
 config SPL_OF_PLATDATA
 	bool "Generate platform data for use in SPL"
 	depends on SPL_OF_CONTROL
scripts/Makefile.lib +0 -0
index 4ea898a421..3b124369ad 100644

scripts/Makefile.lib $(obj)/%.tab.h: $(src)/%.y FORCE · line 242

Quoted 9 lines
scripts/Makefile.lib +2 -1
@@ -235,7 +235,8 @@ $(obj)/%.tab.h: $(src)/%.y FORCE
 # ===========================================================================
 
 quiet_cmd_shipped = SHIPPED $@
-cmd_shipped = cat $< > $@
+cmd_shipped = cat $< | $(objtree)/tools/fdtgrep -r -O dtb - -o $@	\
+		$(addprefix -P ,$(subst $\",,$(CONFIG_OF_REMOVE_PROPS)))
 
 $(obj)/%: $(src)/%_shipped
 	$(call cmd,shipped)

On further thinking, this will make it harder in the case of passing the DTB directly to the kernel. We should guard this option by something else first.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200110/2043dfdd/attachment.sig>
#5 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

Quoted 51 lines · Tom Rini
Quoted 34 lines · Anatolij Gustschin

This can be used for device tree size reduction similar as CONFIG_OF_SPL_REMOVE_PROPS option. For tbs2910 board this shrinks the image size: all -2304.0 bss -16.0 text -2288.0

---
 dts/Kconfig          | 8 ++++++++
 scripts/Makefile.lib | 3 ++-
 2 files changed, 10 insertions(+), 1 deletion(-)
dts/Kconfig +8 -0
index 64c98dd723..49224bee93 100644
@@ -311,6 +311,14 @@ config OF_SPL_REMOVE_PROPS
        can be discarded. This option defines the list of properties to
        discard.
 
+config OF_REMOVE_PROPS
+     string "List of device tree properties to drop"
+     depends on OF_CONTROL
+     default "interrupt-parent interrupts" if PINCTRL
+     help
+       Some properties are not used by U-Boot and can be discarded.
+       This option defines the list of properties to discard.
+
 config SPL_OF_PLATDATA
      bool "Generate platform data for use in SPL"
      depends on SPL_OF_CONTROL
scripts/Makefile.lib +0 -0
index 4ea898a421..3b124369ad 100644

scripts/Makefile.lib $(obj)/%.tab.h: $(src)/%.y FORCE · line 242

Quoted 9 lines
scripts/Makefile.lib +2 -1
@@ -235,7 +235,8 @@ $(obj)/%.tab.h: $(src)/%.y FORCE
 # ===========================================================================
 
 quiet_cmd_shipped = SHIPPED $@
-cmd_shipped = cat $< > $@
+cmd_shipped = cat $< | $(objtree)/tools/fdtgrep -r -O dtb - -o $@    \
+             $(addprefix -P ,$(subst $\",,$(CONFIG_OF_REMOVE_PROPS)))
 
 $(obj)/%: $(src)/%_shipped
      $(call cmd,shipped)

On further thinking, this will make it harder in the case of passing the DTB directly to the kernel. We should guard this option by something else first.

Quoted 2 lines

--
Tom

More importantly, this patch breaks targets that compile pylibfdt, for example, qemu-x86_defconfig.

$ make qemu-x86_defconfig
  HOSTCC  scripts/basic/fixdep
  HOSTCC  scripts/kconfig/conf.o
  YACC    scripts/kconfig/zconf.tab.c
  LEX     scripts/kconfig/zconf.lex.c
  HOSTCC  scripts/kconfig/zconf.tab.o
  HOSTLD  scripts/kconfig/conf
#
# configuration written to .config
#
masahiro at grover:~/ref/u-boot$ make
scripts/kconfig/conf  --syncconfig Kconfig
  CHK     include/config.h
  UPD     include/config.h
  CFG     u-boot.cfg
  GEN     include/autoconf.mk
  GEN     include/autoconf.mk.dep
  CHK     include/config/uboot.release
  UPD     include/config/uboot.release
  CHK     include/generated/version_autogenerated.h
  UPD     include/generated/version_autogenerated.h
  CHK     include/generated/timestamp_autogenerated.h
  UPD     include/generated/timestamp_autogenerated.h
  CC      lib/asm-offsets.s
  CHK     include/generated/generic-asm-offsets.h
  UPD     include/generated/generic-asm-offsets.h
  CC      arch/x86/lib/asm-offsets.s
  CHK     include/generated/asm-offsets.h
  UPD     include/generated/asm-offsets.h
  SHIPPED scripts/dtc/pylibfdt/libfdt.i
/bin/sh: 1: ./tools/fdtgrep: not found
make[3]: *** [scripts/Makefile.lib:242: scripts/dtc/pylibfdt/libfdt.i] Error 127
make[2]: *** [scripts/Makefile.build:432: scripts/dtc/pylibfdt] Error 2
make[1]: *** [scripts/Makefile.build:432: scripts/dtc] Error 2
make: *** [Makefile:551: scripts] Error 2

Please do not touch cmd_shipped.

If this feature is desired,
please implement it in dts/Makefile.

Signature · 2 lines
Best Regards
Masahiro Yamada
#6 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

On Sat, 11 Jan 2020 02:01:27 +0900
Masahiro Yamada masahiroy at kernel.org wrote:
...

Quoted 7 lines

UPD include/generated/asm-offsets.h
SHIPPED scripts/dtc/pylibfdt/libfdt.i
/bin/sh: 1: ./tools/fdtgrep: not found
make[3]: *** [scripts/Makefile.lib:242: scripts/dtc/pylibfdt/libfdt.i] Error 127 make[2]: *** [scripts/Makefile.build:432: scripts/dtc/pylibfdt] Error 2 make[1]: *** [scripts/Makefile.build:432: scripts/dtc] Error 2
make: *** [Makefile:551: scripts] Error 2

Thanks for testing and reporting it, I'll fix it in v2.

Quoted 4 lines

Please do not touch cmd_shipped.

If this feature is desired,
please implement it in dts/Makefile.

OK, thanks!

--
Anatolij

#7 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

In reply to Tom Rini

Hi Tom,

Quoted 11 lines · Tom Rini

scripts/Makefile.lib $(obj)/%.tab.h: $(src)/%.y FORCE · line 242

Quoted 7 lines

quiet_cmd_shipped = SHIPPED $@

scripts/Makefile.lib +2 -1
-cmd_shipped = cat $< > $@
+cmd_shipped = cat $< | $(objtree)/tools/fdtgrep -r -O dtb - -o $@    \
+             $(addprefix -P ,$(subst $\",,$(CONFIG_OF_REMOVE_PROPS)))
 
 $(obj)/%: $(src)/%_shipped
      $(call cmd,shipped)

On further thinking, this will make it harder in the case of passing the DTB directly to the kernel. We should guard this option by something else first.

Please let me ask a question about this.
In my understanding, U-Boot DTB is a different instance from the DTB passed to the kernel.
Was it changed, or is the change planned?

--
Best Regards
Masahiro Yamada

#8 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

Quoted 16 lines · Masahiro Yamada

Hi Tom,

Quoted 11 lines · Tom Rini

scripts/Makefile.lib $(obj)/%.tab.h: $(src)/%.y FORCE · line 242

Quoted 7 lines

quiet_cmd_shipped = SHIPPED $@

scripts/Makefile.lib +2 -1
-cmd_shipped = cat $< > $@
+cmd_shipped = cat $< | $(objtree)/tools/fdtgrep -r -O dtb - -o $@    \
+             $(addprefix -P ,$(subst $\",,$(CONFIG_OF_REMOVE_PROPS)))
 
 $(obj)/%: $(src)/%_shipped
      $(call cmd,shipped)

On further thinking, this will make it harder in the case of passing the DTB directly to the kernel. We should guard this option by something else first.

Quoted 4 lines

Please let me ask a question about this.
In my understanding, U-Boot DTB is a different instance from the DTB passed to the kernel.
Was it changed, or is the change planned?

For a long time now I believe we've had some platforms that wanted to be passed, and pass on again, a single DTB. There's also been boards that have wanted to (or have been) passing what we have on. We also have the case of platforms that more constrained and so stripping information out of the DTB for U-Boot, and continuing to load the kernel DTB from somewhere else, is what they need and want.

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200110/e632b257/attachment.sig>
#9 lore raw

PATCH 1/2 dts: add property removal option CONFIG_OF_REMOVE_PROPS

In reply to Tom Rini

On Fri, 10 Jan 2020 11:39:45 -0500
Tom Rini trini at konsulko.com wrote:
...

Quoted 11 lines

scripts/Makefile.lib $(obj)/%.tab.h: $(src)/%.y FORCE · line 242

Quoted 7 lines

quiet_cmd_shipped = SHIPPED $@

scripts/Makefile.lib +2 -1
-cmd_shipped = cat $< > $@
+cmd_shipped = cat $< | $(objtree)/tools/fdtgrep -r -O dtb - -o $@	\
+		$(addprefix -P ,$(subst $\",,$(CONFIG_OF_REMOVE_PROPS)))
 
 $(obj)/%: $(src)/%_shipped
 	$(call cmd,shipped)  

On further thinking, this will make it harder in the case of passing the DTB directly to the kernel. We should guard this option by something else first.

Agreed, I'll make it optional in patch v2.

--
Anatolij