Skip to contentSkip to conversation

imx: mkimage_fit_atf: replace @ with hyphen

PATCH New Oliver Graute 6 participants 11 messages May 5, 2021 – Aug 16, 2021

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 imx: mkimage_fit_atf: replace @ with hyphen

Unit addresses are not allowed anymore in a FIT image since

commit 3f04db891a35 ("image: Check for unit addresses in FITs")

This caused this error on mkimage creation:

./tools/mkimage: verify_header failed for FIT Image support with exit code 1 Makefile:1393: recipe for target 'u-boot.itb' failed
make: *** [u-boot.itb] Error 1

replacing @ with hyphen solved the issue

Cc Stefano Babic <[email protected]>
Cc Fabio Estevam <[email protected]>
Cc Peng Fan <[email protected]>
Cc Simon Glass <[email protected]>
Cc uboot-imx <[email protected]>
---
 arch/arm/mach-imx/mkimage_fit_atf.sh | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
arch/arm/mach-imx/mkimage_fit_atf.sh +12 -12
index fe12b7bb4b..c5fb9b0fc3 100755
@@ -53,7 +53,7 @@ cat << __HEADER_EOF
 	description = "Configuration to load ATF before U-Boot";
 
 	images {
-		uboot at 1 {
+		uboot-1 {
 			description = "U-Boot (64-bit)";
 			os = "u-boot";
 			data = /incbin/("$BL33");
@@ -68,7 +68,7 @@ cnt=1
 for dtname in $*
 do
 	cat << __FDT_IMAGE_EOF
-		fdt@$cnt {
+		fdt-$cnt {
 			description = "$(basename $dtname .dtb)";
 			data = /incbin/("$dtname");
 			type = "flat_dt";
@@ -79,7 +79,7 @@ cnt=$((cnt+1))
 done
 
 cat << __HEADER_EOF
-		atf at 1 {
+		atf-1 {
 			description = "ARM Trusted Firmware";
 			os = "arm-trusted-firmware";
 			data = /incbin/("$BL31");
@@ -93,7 +93,7 @@ __HEADER_EOF
 
 if [ -f $BL32 ]; then
 cat << __HEADER_EOF
-		tee at 1 {
+		tee-1 {
 			description = "TEE firmware";
 			data = /incbin/("$BL32");
 			type = "firmware";
@@ -108,7 +108,7 @@ fi
 cat << __CONF_HEADER_EOF
 	};
 	configurations {
-		default = "config at 1";
+		default = "config-1";
 
 __CONF_HEADER_EOF
 
@@ -117,20 +117,20 @@ for dtname in $*
 do
 if [ -f $BL32 ]; then
 cat << __CONF_SECTION_EOF
-		config@$cnt {
+		config-$cnt {
 			description = "$(basename $dtname .dtb)";
-			firmware = "uboot at 1";
-			loadables = "atf at 1", "tee at 1";
+			firmware = "uboot-1";
+			loadables = "atf-1", "tee-1";
 			fdt = "fdt@$cnt";
 		};
 __CONF_SECTION_EOF
 else
 cat << __CONF_SECTION1_EOF
-		config@$cnt {
+		config-$cnt {
 			description = "$(basename $dtname .dtb)";
-			firmware = "uboot at 1";
-			loadables = "atf@1";
-			fdt = "fdt@$cnt";
+			firmware = "uboot-1";
+			loadables = "atf-1";
+			fdt = "fdt-$cnt";
 		};
 __CONF_SECTION1_EOF
 fi
Signature · 1 line
2.17.1
#1 lore raw

PATCH 1/1 imx: mkimage_fit_atf: replace @ with hyphen

Hello Oliver,

Quoted 12 lines

-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Oliver Graute
Sent: Wednesday, May 5, 2021 2:01 PM
To: sbabic at denx.de
Cc: festevam at gmail.com; peng.fan at nxp.com; oliver.graute at kococonnector.com; u-boot at lists.denx.de; Simon Glass <[email protected]>; uboot-imx <[email protected]>
Subject: [PATCH v1] imx: mkimage_fit_atf: replace @ with hyphen

Unit addresses are not allowed anymore in a FIT image since

Quoted 7 lines

commit 3f04db891a35 ("image: Check for unit addresses in FITs")

This caused this error on mkimage creation:

./tools/mkimage: verify_header failed for FIT Image support with exit code 1 Makefile:1393: recipe for target 'u-boot.itb' failed
make: *** [u-boot.itb] Error 1

IIRC, this error comes from the derivatives which are not (yet) migrated to use 'binman' and have CONFIG_SPL_FIT_GENERATOR set.

It is triggered by 79af75f777 ("fit: Don't allow verification of images with @ nodes"), which is totally correct behavior and should be corrected on the FIT generator end, so the patch itself solves the problem.

TBH, I had introduced the same change to NXP U-Boot fork that is built for NXP Yocto BSP [1], and was not sure it is relevant for the master since it uses 'binman' instead of FIT generator.

From the other point of view, which imx8m board do you experience this build problem with?

Currently, there are following boards that need to be migrated from FIT generator to 'binman':
configs/imx8mm_beacon_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mm_venice_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mn_beacon_2g_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mn_beacon_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mq_evk_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8mq_phanbell_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/imx8qm_rom7720_a1_4G_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/phycore-imx8mm_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/phycore-imx8mp_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/pico-imx8mq_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh" configs/verdin-imx8mm_defconfig:CONFIG_SPL_FIT_GENERATOR="arch/arm/mach-imx/mkimage_fit_atf.sh"

I know it's a long shot, but the question is: should the FIT generator be adapted here, or maybe those boards that are still using it - to be migrated to binman directly?

There is already a warning in root Makefile regarding usage of FIT generator:
"===================== WARNING ======================"
"This board uses CONFIG_SPL_FIT_GENERATOR. Please migrate"
"to binman instead, to avoid the proliferation of"
"arch-specific scripts with no tests." "===================================================="

So maybe it makes sense to adapt affected boards instead?

Quoted 86 lines

replacing @ with hyphen solved the issue

Cc Stefano Babic <[email protected]>
Cc Fabio Estevam <[email protected]>
Cc Peng Fan <[email protected]>
Cc Simon Glass <[email protected]>
Cc uboot-imx <[email protected]>
---
 arch/arm/mach-imx/mkimage_fit_atf.sh | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
arch/arm/mach- renamed from arch/arm/mach-imx/mkimage_fit_atf.sh +0 -0

imx/mkimage_fit_atf.sh
index fe12b7bb4b..c5fb9b0fc3 100755

arch/arm/mach-imx/mkimage_fit_atf.sh +1 -1
@@ -53,7 +53,7 @@ cat << __HEADER_EOF
        description = "Configuration to load ATF before U-Boot";
 
        images {
-               uboot at 1 {
+               uboot-1 {
                        description = "U-Boot (64-bit)";
                        os = "u-boot";
                        data = /incbin/("$BL33"); @@ -68,7 +68,7 @@ cnt=1  for dtname in $*

do

        cat << __FDT_IMAGE_EOF
arch/arm/mach-imx/mkimage_fit_atf.sh +3 -3
-               fdt@$cnt {
+               fdt-$cnt {
                        description = "$(basename $dtname .dtb)";
                        data = /incbin/("$dtname");
                        type = "flat_dt"; @@ -79,7 +79,7 @@ cnt=$((cnt+1))  done
 
 cat << __HEADER_EOF
-               atf at 1 {
+               atf-1 {
                        description = "ARM Trusted Firmware";
                        os = "arm-trusted-firmware";
                        data = /incbin/("$BL31"); @@ -93,7 +93,7 @@ __HEADER_EOF
 
 if [ -f $BL32 ]; then
 cat << __HEADER_EOF
-               tee at 1 {
+               tee-1 {
                        description = "TEE firmware";
                        data = /incbin/("$BL32");
                        type = "firmware"; @@ -108,7 +108,7 @@ fi  cat <<

__CONF_HEADER_EOF

        };
        configurations {
arch/arm/mach-imx/mkimage_fit_atf.sh +8 -9
-               default = "config at 1";
+               default = "config-1";
 
 __CONF_HEADER_EOF
@@ -117,20 +117,20 @@ for dtname in $*
 do
 if [ -f $BL32 ]; then
 cat << __CONF_SECTION_EOF
-               config@$cnt {
+               config-$cnt {
                        description = "$(basename $dtname .dtb)";
-                       firmware = "uboot at 1";
-                       loadables = "atf at 1", "tee at 1";
+                       firmware = "uboot-1";
+                       loadables = "atf-1", "tee-1";
                        fdt = "fdt@$cnt";
                };
 __CONF_SECTION_EOF
 else
 cat << __CONF_SECTION1_EOF
-               config@$cnt {
+               config-$cnt {
                        description = "$(basename $dtname .dtb)";
-                       firmware = "uboot at 1";
-                       loadables = "atf at 1";
-                       fdt = "fdt@$cnt";
+                       firmware = "uboot-1";
+                       loadables = "atf-1";
+                       fdt = "fdt-$cnt";
                };
 __CONF_SECTION1_EOF
 fi
--
Quoted 1 line

2.17.1

-- andrey

#2 lore raw

Hi all,

Am Mi., 5. Mai 2021 um 15:30 Uhr schrieb ZHIZHIKIN Andrey <[email protected]>:

Quoted 22 lines

Hello Oliver,

Quoted 19 lines

-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Oliver Graute
Sent: Wednesday, May 5, 2021 2:01 PM
To: [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; Simon Glass <[email protected]>; uboot-imx <[email protected]>
Subject: [PATCH v1] imx: mkimage_fit_atf: replace @ with hyphen

Unit addresses are not allowed anymore in a FIT image since

commit 3f04db891a35 ("image: Check for unit addresses in FITs")

This caused this error on mkimage creation:

./tools/mkimage: verify_header failed for FIT Image support with exit code 1 Makefile:1393: recipe for target 'u-boot.itb' failed
make: *** [u-boot.itb] Error 1

This error is still present in the current master branch and the target "flash.bin" does not not build anymore for the imx8mq_evk_defconfig.

Any thoughts to apply this patch to master?

--
Heiko

#3 lore raw
Quoted 28 lines · Heiko Thiery

Hi all,

Am Mi., 5. Mai 2021 um 15:30 Uhr schrieb ZHIZHIKIN Andrey <[email protected]>:

Quoted 22 lines

Hello Oliver,

Quoted 19 lines

-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Oliver Graute
Sent: Wednesday, May 5, 2021 2:01 PM
To: [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; Simon Glass <[email protected]>; uboot-imx <[email protected]>
Subject: [PATCH v1] imx: mkimage_fit_atf: replace @ with hyphen

Unit addresses are not allowed anymore in a FIT image since

commit 3f04db891a35 ("image: Check for unit addresses in FITs")

This caused this error on mkimage creation:

./tools/mkimage: verify_header failed for FIT Image support with exit code 1 Makefile:1393: recipe for target 'u-boot.itb' failed
make: *** [u-boot.itb] Error 1

Quoted 6 lines

This error is still present in the current master branch and the target "flash.bin" does not not build anymore for the imx8mq_evk_defconfig.

Any thoughts to apply this patch to master?

Heiko,

I just ran into this failure as well with imx8mm_venice_defconfig which I haven't moved to binman yet.

With respect to moving to binman, I'm all for it but I'm wondering why other boards have elected to move from a monolithic flash.bin including the SPL and u-boot.itb pre-binman to post-binman where flash.bin is just the SPL. I'm not sure why users should be hit with a change like that... why are people not having binman produce a flash.bin that was equivalent to pre-binman?

Best regards,

Tim

#4 lore raw

Hi Tim,

Quoted 43 lines

Subject: Re: [PATCH v1] imx: mkimage_fit_atf: replace @ with hyphen

Quoted 30 lines · Heiko Thiery

Hi all,

Am Mi., 5. Mai 2021 um 15:30 Uhr schrieb ZHIZHIKIN Andrey <[email protected]>:

Quoted 24 lines

Hello Oliver,

Quoted 21 lines

-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Oliver Graute
Sent: Wednesday, May 5, 2021 2:01 PM
To: [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; Simon Glass <[email protected]>; uboot-imx <[email protected]>
Subject: [PATCH v1] imx: mkimage_fit_atf: replace @ with hyphen

Unit addresses are not allowed anymore in a FIT image since

commit 3f04db891a35 ("image: Check for unit addresses in FITs")

This caused this error on mkimage creation:

./tools/mkimage: verify_header failed for FIT Image support with exit code 1
Makefile:1393: recipe for target 'u-boot.itb' failed
make: *** [u-boot.itb] Error 1

Quoted 6 lines

This error is still present in the current master branch and the target "flash.bin" does not not build anymore for the imx8mq_evk_defconfig.

Any thoughts to apply this patch to master?

Heiko,

Quoted 8 lines

I just ran into this failure as well with imx8mm_venice_defconfig which I haven't moved to binman yet.

With respect to moving to binman, I'm all for it but I'm wondering why other boards have elected to move from a monolithic flash.bin including the SPL and u-boot.itb pre-binman to post-binman where flash.bin is just the SPL. I'm not sure why users should be hit with a change like that... why are people not having binman produce a flash.bin that was equivalent to pre-binman?

You mean flash.bin not contain u-boot.bin when using binman here?

Regards,
Peng.

Quoted 4 lines

Best regards,

Tim

#5 lore raw

Hi all,

Am Mo., 2. Aug. 2021 um 11:57 Uhr schrieb Peng Fan (OSS) <[email protected]>:

Quoted 49 lines

Hi Tim,

Quoted 46 lines

Subject: Re: [PATCH v1] imx: mkimage_fit_atf: replace @ with hyphen

Quoted 30 lines · Heiko Thiery

Hi all,

Am Mi., 5. Mai 2021 um 15:30 Uhr schrieb ZHIZHIKIN Andrey <[email protected]>:

Quoted 24 lines

Hello Oliver,

Quoted 21 lines

-----Original Message-----
From: U-Boot <[email protected]> On Behalf Of Oliver Graute
Sent: Wednesday, May 5, 2021 2:01 PM
To: [email protected]
Cc: [email protected]; [email protected]; [email protected]; [email protected]; Simon Glass <[email protected]>; uboot-imx <[email protected]>
Subject: [PATCH v1] imx: mkimage_fit_atf: replace @ with hyphen

Unit addresses are not allowed anymore in a FIT image since

commit 3f04db891a35 ("image: Check for unit addresses in FITs")

This caused this error on mkimage creation:

./tools/mkimage: verify_header failed for FIT Image support with exit code 1
Makefile:1393: recipe for target 'u-boot.itb' failed
make: *** [u-boot.itb] Error 1

Quoted 6 lines

This error is still present in the current master branch and the target "flash.bin" does not not build anymore for the imx8mq_evk_defconfig.

Any thoughts to apply this patch to master?

Heiko,

I just ran into this failure as well with imx8mm_venice_defconfig which I haven't moved to binman yet.

Quoted 7 lines
Quoted 5 lines

With respect to moving to binman, I'm all for it but I'm wondering why other boards have elected to move from a monolithic flash.bin including the SPL and u-boot.itb pre-binman to post-binman where flash.bin is just the SPL. I'm not sure why users should be hit with a change like that... why are people not having binman produce a flash.bin that was equivalent to pre-binman?

You mean flash.bin not contain u-boot.bin when using binman here?

Any news on that? I just see that in the current master it is not possible to create a working flash.bin for e.g. the imx8mq_evk board.

Signature · 1 line
Heiko
reviewer
#6 lore raw
In reply to Tim Harvey

Hi Tim,

[Adding Marek]

Quoted 6 lines · Tim Harvey

With respect to moving to binman, I'm all for it but I'm wondering why other boards have elected to move from a monolithic flash.bin including the SPL and u-boot.itb pre-binman to post-binman where flash.bin is just the SPL. I'm not sure why users should be hit with a change like that... why are people not having binman produce a flash.bin that was equivalent to pre-binman?

Yes, fully agree.

Peng,

After the adoption of binman, both flash.bin and u-boot.itb need to be flashed to the boot media.

This breaks user experience, distro integration, etc, which is not good.

Can we have binman generate a single bootable flash.bin binary (which contains flash.bin + u-boot.itb) again
to keep compatibility and avoid breakage?

Thanks,

Fabio Estevam

Quoted 4 lines

Best regards,

Tim

#7 lore raw

Hi,

Am Fr., 13. Aug. 2021 um 14:34 Uhr schrieb Fabio Estevam <[email protected]>:

Quoted 18 lines

Hi Tim,

[Adding Marek]

Quoted 6 lines · Tim Harvey

With respect to moving to binman, I'm all for it but I'm wondering why other boards have elected to move from a monolithic flash.bin including the SPL and u-boot.itb pre-binman to post-binman where flash.bin is just the SPL. I'm not sure why users should be hit with a change like that... why are people not having binman produce a flash.bin that was equivalent to pre-binman?

Yes, fully agree.

Peng,

Quoted 8 lines

After the adoption of binman, both flash.bin and u-boot.itb need to be flashed to the boot media.

This breaks user experience, distro integration, etc, which is not good.

Can we have binman generate a single bootable flash.bin binary (which contains flash.bin + u-boot.itb) again
to keep compatibility and avoid breakage?

Maybe I'm wrong or I do not fully understand the limitation you're talking about, but as far as I understand the output (flash.bin) from the ronetix board [1] was generated using binman and includes all necessary parts. Is this right?

[1] https://source.denx.de/u-boot/u-boot/-/blob/master/arch/arm/dts/imx8mq-cm-u-boot.dtsi

Signature · 1 line
Heiko
reviewer
#8 lore raw

Hi Heiko,

Quoted 4 lines · Heiko Thiery

Maybe I'm wrong or I do not fully understand the limitation you're talking about, but as far as I understand the output (flash.bin) from the ronetix board [1] was generated using binman and includes all necessary parts. Is this right?

Building the imx8mq_cm_defconfig generates flash.bin + u-boot.itb.

If you only flash flash.bin, then you don't have a bootable system as flash.bin contains only the SPL part.

Prior to using binman, we could boot the i.MX8M boards by flashing only flash.bin.

I would like to preserver the pre-U-Boot 2021.07, where only flash.bin was needed.

Quoted 6 lines

[1] https://source.denx.de/u-boot/u-boot/-/blob/master/arch/arm/dts/imx8mq-cm-u-boot.dtsi

--
Heiko

#9 lore raw

Hi Fabio,

Am Fr., 13. Aug. 2021 um 15:03 Uhr schrieb Fabio Estevam <[email protected]>:

Quoted 12 lines

Hi Heiko,

Quoted 4 lines · Heiko Thiery

Maybe I'm wrong or I do not fully understand the limitation you're talking about, but as far as I understand the output (flash.bin) from the ronetix board [1] was generated using binman and includes all necessary parts. Is this right?

Building the imx8mq_cm_defconfig generates flash.bin + u-boot.itb.

Quoted 8 lines

If you only flash flash.bin, then you don't have a bootable system as flash.bin contains only the SPL part.

Prior to using binman, we could boot the i.MX8M boards by flashing only flash.bin.

I would like to preserver the pre-U-Boot 2021.07, where only flash.bin was needed.

I had taken the binman configuration from imx8mq_cm_defconfig and adapted it for my board (kontron-pitx-imx8m). I had copied the flash.bin image to my SD card and am of the opinion that this had started afterwards with the current u-boot. Unfortunately I can test it only next week again, because I currently have no device at hand.

I must confess that I do not yet fully understand how binman works.

I'll get back to you next week.

Signature · 1 line
Heiko
#10 lore raw

HI Fabrio and all,

Am Fr., 13. Aug. 2021 um 17:12 Uhr schrieb Heiko Thiery <[email protected]>:

Quoted 22 lines

Hi Fabio,

Am Fr., 13. Aug. 2021 um 15:03 Uhr schrieb Fabio Estevam <[email protected]>:

Quoted 18 lines

Hi Heiko,

Quoted 4 lines · Heiko Thiery

Maybe I'm wrong or I do not fully understand the limitation you're talking about, but as far as I understand the output (flash.bin) from the ronetix board [1] was generated using binman and includes all necessary parts. Is this right?

Building the imx8mq_cm_defconfig generates flash.bin + u-boot.itb.

If you only flash flash.bin, then you don't have a bootable system as flash.bin contains only the SPL part.

Prior to using binman, we could boot the i.MX8M boards by flashing only flash.bin.

Quoted 8 lines
Quoted 2 lines

I would like to preserver the pre-U-Boot 2021.07, where only flash.bin was needed.

I had taken the binman configuration from imx8mq_cm_defconfig and adapted it for my board (kontron-pitx-imx8m). I had copied the flash.bin image to my SD card and am of the opinion that this had started afterwards with the current u-boot. Unfortunately I can test it only next week again, because I currently have no device at hand.

You were right. I probably still had the old data from u-boot.itb on the SD card and therefore the board still started correctly.

But I also think that the flash.bin should have the complete content (spl + u-boot.itb) so that the board starts correctly.

Sorry for the noise.

Signature · 1 line
Heiko