Initial support for NXP iMX8 (imx8mq-evk)

This package add support for imx-mkimage for iMX8M. Currently the only hardware it's tested for.

Signed-off-by Erik Larsson <[email protected]>

Signed-off-by Christopher Dahlberg <[email protected]>

 package/Config.in.host                             |  1 +
 ...-for-overriding-bl32-and-bl33-not-only-bl.patch | 58 ++++++++++++++++++++++
 .../0002-add-ldflags-to-link-step.patch            | 26 ++++++++++
 package/imx-mkimage/Config.in.host                 |  7 +++
 package/imx-mkimage/imx-mkimage.hash               |  3 ++
 package/imx-mkimage/imx-mkimage.mk                 | 24 +++++++++
 6 files changed, 119 insertions(+)

Hello,

Again, thanks for this contribution!

Quoted 5 lines · Erik Larsson

This package add support for imx-mkimage for iMX8M. Currently the only hardware it's tested for.

Signed-off-by Erik Larsson <[email protected]>

Signed-off-by Christopher Dahlberg <[email protected]>

Thanks, I've applied to master, after doing a few changes. I also have one request. See below.

Quoted 7 lines

 package/Config.in.host                             |  1 +
 ...-for-overriding-bl32-and-bl33-not-only-bl.patch | 58 ++++++++++++++++++++++
 .../0002-add-ldflags-to-link-step.patch            | 26 ++++++++++
 package/imx-mkimage/Config.in.host                 |  7 +++
 package/imx-mkimage/imx-mkimage.hash               |  3 ++
 package/imx-mkimage/imx-mkimage.mk                 | 24 +++++++++
 6 files changed, 119 insertions(+)

You forgot to add an entry in the DEVELOPERS file for this new package, so I've done so.

Quoted 9 lines

package/imx-mkimage/0002-add-ldflags-to-link-step.patch new file +3 -0

new file mode 100644
index 0000000..22c6a7d
@@ -0,0 +1,26 @@
+From a73f20fbe921c7ecff3efda23bc506fdd935ee4d Mon Sep 17 00:00:00 2001
+From: Erik Larsson <[email protected]>
+Date: Tue, 6 Mar 2018 12:28:39 +0100

package/imx-mkimage/0002-add-ldflags-to-link-step.patch line 6

Quoted 3 lines

package/imx-mkimage/0002-add-ldflags-to-link-step.patch new file +3 -0

+Subject: [PATCH 1/2] Add LDFLAGS to link step
+
+Signed-off-by: Erik Larsson <[email protected]>

This commit log really needs to be extended. What was the need to add LDFLAGS ? It seems like a very simple program, which could be built without special LDFLAGS. What was the issue that lead you to add this patch ?

package/imx-mkimage/0002-add-ldflags-to-link-step.patch line 9

Quoted 5 lines

package/imx-mkimage/0002-add-ldflags-to-link-step.patch new file +3 -0

+define HOST_IMX_MKIMAGE_INSTALL_CMDS
+     $(INSTALL) -D -m 755 $(@D)/iMX8M/mkimage_imx8 $(HOST_DIR)/bin
+     $(INSTALL) -D -m 755 $(@D)/iMX8M/mkimage_fit_atf.sh $(HOST_DIR)/bin

We prefer a complete destination path that includes the name of the file to install. Indeed, with the way you have written the commands, if $(HOST_DIR)/bin doesn't exist as a directory, mkimage_imx8 will be installed as a file called "bin" in $(HOST_DIR). I've also fixed this when applying.

Thanks!

Thomas
--
Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons) Embedded Linux and Kernel engineering
https://bootlin.com

Quoted 4 lines

This commit log really needs to be extended. What was the need to add LDFLAGS ? It seems like a very simple program, which could be built without special LDFLAGS. What was the issue that lead you to add this patch ?

If LDFLAGS is not added it will use the host machine libz, and fail if not available, but by adding LDFLAGS it will build using Buildroot compiled libz.

Do you want me to update the patch information in a new patch?

Signed-off-by Erik Larsson <[email protected]>

Signed-off-by Christopher Dahlberg <[email protected]>

---
v6: No changes
v5: No changes
v4: Changed title on commit (suggested by Gary B.)

---
 package/freescale-imx/Config.in | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

board/freescale/common/imx/post-image.sh +2 -2

index b239743..4c77ae2 100755
@@ -7,10 +7,10 @@
 #
 dtb_list()
 {
-	local DTB_LIST="$(sed -n 's/^BR2_LINUX_KERNEL_INTREE_DTS_NAME="\([a-z0-9 \-]*\)"$/\1/p' ${BR2_CONFIG})"
+	local DTB_LIST="$(sed -n 's/^BR2_LINUX_KERNEL_INTREE_DTS_NAME="\([\/a-z0-9 \-]*\)"$/\1/p' ${BR2_CONFIG})"
 
 	for dt in $DTB_LIST; do
-		echo -n "\"$dt.dtb\", "
+		echo -n "\"`basename $dt`.dtb\", "
 	done
 }
 

board/freescale/common/imx/post-image.sh +6 -1

@@ -30,19 +32,39 @@ linux_image()
 
 genimage_type()
 {
-	if grep -Eq "^BR2_TARGET_UBOOT_SPL=y$" ${BR2_CONFIG}; then
+	if grep -Eq "^BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX8M=y$" ${BR2_CONFIG}; then
+		echo "genimage.cfg.template_imx8"
+	elif grep -Eq "^BR2_TARGET_UBOOT_SPL=y$" ${BR2_CONFIG}; then
 		echo "genimage.cfg.template_spl"
 	else
 		echo "genimage.cfg.template"
 	fi
 }
 
+gen_imx8_boot()
+{
+	cat ${BINARIES_DIR}/lpddr4_pmu_train_1d_imem_pad.bin ${BINARIES_DIR}/lpddr4_pmu_train_1d_dmem_pad.bin > ${BINARIES_DIR}/lpddr4_pmu_train_1d_fw.bin

Shouldn't this logic be moved to a specialized post-image script ? And then you do:

BR2_ROOTFS_POST_IMAGE_SCRIPT="board/freescale/common/imx/imx8-bootloader-prepare.sh board/freescale/common/imx/post-image.sh"

or something like that ? Fabio, what do you think ?

Hello Erik,

Quoted 18 lines · Erik Larsson

board/freescale/common/imx/post-image.sh +2 -2

index b239743..4c77ae2 100755
@@ -7,10 +7,10 @@
 #
 dtb_list()
 {
-     local DTB_LIST="$(sed -n 's/^BR2_LINUX_KERNEL_INTREE_DTS_NAME="\([a-z0-9 \-]*\)"$/\1/p' ${BR2_CONFIG})"
+     local DTB_LIST="$(sed -n 's/^BR2_LINUX_KERNEL_INTREE_DTS_NAME="\([\/a-z0-9 \-]*\)"$/\1/p' ${BR2_CONFIG})"
 
      for dt in $DTB_LIST; do
-             echo -n "\"$dt.dtb\", "
+             echo -n "\"`basename $dt`.dtb\", "
      done
 }
 

board/freescale/common/imx/post-image.sh linux_image() · line 30

Quoted 8 lines

board/freescale/common/imx/post-image.sh +2 -0

@@ -23,6 +23,8 @@ linux_image()
 {
      if grep -Eq "^BR2_LINUX_KERNEL_UIMAGE=y$" ${BR2_CONFIG}; then
              echo "\"uImage\""
+     elif grep -Eq "^BR2_LINUX_KERNEL_IMAGE=y$" ${BR2_CONFIG}; then
+             echo "\"Image\""
      else
              echo "\"zImage\""
      fi

Those two changes could be in a patch reworking post-image.sh to be compatible with arm64.

Quoted 17 lines

board/freescale/common/imx/post-image.sh +6 -1

@@ -30,19 +32,39 @@ linux_image()
 
 genimage_type()
 {
-     if grep -Eq "^BR2_TARGET_UBOOT_SPL=y$" ${BR2_CONFIG}; then
+     if grep -Eq "^BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX8M=y$" ${BR2_CONFIG}; then
+             echo "genimage.cfg.template_imx8"
+     elif grep -Eq "^BR2_TARGET_UBOOT_SPL=y$" ${BR2_CONFIG}; then
              echo "genimage.cfg.template_spl"
      else
              echo "genimage.cfg.template"
      fi
 }
 
+gen_imx8_boot()
+{
+     cat ${BINARIES_DIR}/lpddr4_pmu_train_1d_imem_pad.bin ${BINARIES_DIR}/lpddr4_pmu_train_1d_dmem_pad.bin > ${BINARIES_DIR}/lpddr4_pmu_train_1d_fw.bin

board/freescale/common/imx/post-image.sh linux_image() · line 56

Quoted 10 lines

board/freescale/common/imx/post-image.sh +10 -0

+     cat ${BINARIES_DIR}/lpddr4_pmu_train_2d_imem_pad.bin ${BINARIES_DIR}/lpddr4_pmu_train_2d_dmem.bin > ${BINARIES_DIR}/lpddr4_pmu_train_2d_fw.bin
+     cat ${BINARIES_DIR}/u-boot-spl.bin ${BINARIES_DIR}/lpddr4_pmu_train_1d_fw.bin ${BINARIES_DIR}/lpddr4_pmu_train_2d_fw.bin > ${BINARIES_DIR}/u-boot-spl-ddr.bin
+     rm -f ${BINARIES_DIR}/lpddr4_pmu_train_1d_fw.bin ${BINARIES_DIR}/lpddr4_pmu_train_2d_fw.bin
+
+     BL31=${BINARIES_DIR}/bl31.bin BL33=${BINARIES_DIR}/u-boot.bin ${HOST_DIR}/bin/mkimage_fit_atf.sh ${BINARIES_DIR}/fsl-imx8mq-evk.dtb > ${BINARIES_DIR}/u-boot.its
+     ${HOST_DIR}/bin/mkimage -E -p 0x3000 -f ${BINARIES_DIR}/u-boot.its ${BINARIES_DIR}/u-boot.itb
+     rm -f ${BINARIES_DIR}/u-boot.its
+
+     ${HOST_DIR}/bin/mkimage_imx8 -fit -signed_hdmi ${BINARIES_DIR}/signed_hdmi_imx8m.bin -loader ${BINARIES_DIR}/u-boot-spl-ddr.bin 0x7E1000 -second_loader ${BINARIES_DIR}/u-boot.itb 0x40200000 0x60000 -out ${BINARIES_DIR}/imx-boot-imx8mqevk-sd.bin
+}

Shouldn't this logic be moved to a specialized post-image script ? And then you do:

BR2_ROOTFS_POST_IMAGE_SCRIPT="board/freescale/common/imx/imx8-bootloader-prepare.sh board/freescale/common/imx/post-image.sh"

or something like that ? Fabio, what do you think ?

Also, could you split the i.MX8 stuff from the support of the i.MX8 EVK board itself ?

I.e adding the genimage.cfg file, and adding imx8-bootloader.prepare.sh should be one patch, and then adding the i.MX8 EVK defconfig and readme.txt should be another patch.

Could you rework the last two patches according to those comments ?

Thanks a lot!

Thomas Petazzoni
--
Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons) Embedded Linux and Kernel engineering
https://bootlin.com

Hello Erik,

Quoted 18 lines · Erik Larsson

board/freescale/common/imx/post-image.sh +2 -2

index b239743..4c77ae2 100755
@@ -7,10 +7,10 @@
 #
 dtb_list()
 {
-     local DTB_LIST="$(sed -n 's/^BR2_LINUX_KERNEL_INTREE_DTS_NAME="\([a-z0-9 \-]*\)"$/\1/p' ${BR2_CONFIG})"
+     local DTB_LIST="$(sed -n 's/^BR2_LINUX_KERNEL_INTREE_DTS_NAME="\([\/a-z0-9 \-]*\)"$/\1/p' ${BR2_CONFIG})"
 
      for dt in $DTB_LIST; do
-             echo -n "\"$dt.dtb\", "
+             echo -n "\"`basename $dt`.dtb\", "
      done
 }
 

board/freescale/common/imx/post-image.sh linux_image() · line 30

Quoted 8 lines

board/freescale/common/imx/post-image.sh +2 -0

@@ -23,6 +23,8 @@ linux_image()
 {
      if grep -Eq "^BR2_LINUX_KERNEL_UIMAGE=y$" ${BR2_CONFIG}; then
              echo "\"uImage\""
+     elif grep -Eq "^BR2_LINUX_KERNEL_IMAGE=y$" ${BR2_CONFIG}; then
+             echo "\"Image\""
      else
              echo "\"zImage\""
      fi

Those two changes could be in a patch reworking post-image.sh to be compatible with arm64.

Quoted 17 lines

board/freescale/common/imx/post-image.sh +6 -1

@@ -30,19 +32,39 @@ linux_image()
 
 genimage_type()
 {
-     if grep -Eq "^BR2_TARGET_UBOOT_SPL=y$" ${BR2_CONFIG}; then
+     if grep -Eq "^BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX8M=y$" ${BR2_CONFIG}; then
+             echo "genimage.cfg.template_imx8"
+     elif grep -Eq "^BR2_TARGET_UBOOT_SPL=y$" ${BR2_CONFIG}; then
              echo "genimage.cfg.template_spl"
      else
              echo "genimage.cfg.template"
      fi
 }
 
+gen_imx8_boot()
+{
+     cat ${BINARIES_DIR}/lpddr4_pmu_train_1d_imem_pad.bin ${BINARIES_DIR}/lpddr4_pmu_train_1d_dmem_pad.bin > ${BINARIES_DIR}/lpddr4_pmu_train_1d_fw.bin

board/freescale/common/imx/post-image.sh linux_image() · line 56

Quoted 10 lines

board/freescale/common/imx/post-image.sh +10 -0

+     cat ${BINARIES_DIR}/lpddr4_pmu_train_2d_imem_pad.bin ${BINARIES_DIR}/lpddr4_pmu_train_2d_dmem.bin > ${BINARIES_DIR}/lpddr4_pmu_train_2d_fw.bin
+     cat ${BINARIES_DIR}/u-boot-spl.bin ${BINARIES_DIR}/lpddr4_pmu_train_1d_fw.bin ${BINARIES_DIR}/lpddr4_pmu_train_2d_fw.bin > ${BINARIES_DIR}/u-boot-spl-ddr.bin
+     rm -f ${BINARIES_DIR}/lpddr4_pmu_train_1d_fw.bin ${BINARIES_DIR}/lpddr4_pmu_train_2d_fw.bin
+
+     BL31=${BINARIES_DIR}/bl31.bin BL33=${BINARIES_DIR}/u-boot.bin ${HOST_DIR}/bin/mkimage_fit_atf.sh ${BINARIES_DIR}/fsl-imx8mq-evk.dtb > ${BINARIES_DIR}/u-boot.its
+     ${HOST_DIR}/bin/mkimage -E -p 0x3000 -f ${BINARIES_DIR}/u-boot.its ${BINARIES_DIR}/u-boot.itb
+     rm -f ${BINARIES_DIR}/u-boot.its
+
+     ${HOST_DIR}/bin/mkimage_imx8 -fit -signed_hdmi ${BINARIES_DIR}/signed_hdmi_imx8m.bin -loader ${BINARIES_DIR}/u-boot-spl-ddr.bin 0x7E1000 -second_loader ${BINARIES_DIR}/u-boot.itb 0x40200000 0x60000 -out ${BINARIES_DIR}/imx-boot-imx8mqevk-sd.bin
+}

Shouldn't this logic be moved to a specialized post-image script ? And then you do:

BR2_ROOTFS_POST_IMAGE_SCRIPT="board/freescale/common/imx/imx8-bootloader-prepare.sh board/freescale/common/imx/post-image.sh"

or something like that ? Fabio, what do you think ?

Also, could you split the i.MX8 stuff from the support of the i.MX8 EVK board itself ?

I.e adding the genimage.cfg file, and adding imx8-bootloader.prepare.sh should be one patch, and then adding the i.MX8 EVK defconfig and readme.txt should be another patch.

Could you rework the last two patches according to those comments ?

Thanks a lot!

Thomas Petazzoni
--
Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons) Embedded Linux and Kernel engineering
https://bootlin.com

Quoted 7 lines

Also, could you split the i.MX8 stuff from the support of the i.MX8 EVK board itself ?

I.e adding the genimage.cfg file, and adding imx8-bootloader.prepare.sh should be one patch, and then adding the i.MX8 EVK defconfig and readme.txt should be another patch.

Could you rework the last two patches according to those comments ?

I've started to rework this and have something that at least works. I've moved the generation of imx-boot-imx8mqevk-sd.bin to a separate prepare script and tried to separate what is iMX8 general and what is imx8mqevk specific in that script. So the gen_imx8_boot() function can be removed from post-script.sh. I still need to keep the selection of genimage.cfg.template in post-script.sh but instead of checking for IMX8M it can check if it's a IMX8 platform. Is it something like that you hade in mind?

Hello,

Quoted 8 lines · Erik Larsson

Quoted 7 lines

Also, could you split the i.MX8 stuff from the support of the i.MX8 EVK board itself ?

I.e adding the genimage.cfg file, and adding imx8-bootloader.prepare.sh should be one patch, and then adding the i.MX8 EVK defconfig and readme.txt should be another patch.

Could you rework the last two patches according to those comments ?

Quoted 8 lines

I've started to rework this and have something that at least works. I've moved the generation of imx-boot-imx8mqevk-sd.bin to a separate prepare script and tried to separate what is iMX8 general and what is imx8mqevk specific in that script. So the gen_imx8_boot() function can be removed from post-script.sh. I still need to keep the selection of genimage.cfg.template in post-script.sh but instead of checking for IMX8M it can check if it's a IMX8 platform. Is it something like that you hade in mind?

How are you going to check if it's an IMX8 platform ? For me, the test on BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX8M to decide which genimage file to use was fine. What I wanted is:

(1) Split in different commits the introduction of i.MX8 support and the addition of the i.MX8 EVK support. A first commit should change the post-image.sh script to add i.MX8 stuff, and separate commit should add the defconfig and board/ stuff related to the i.MX8 EVK board.

(2) Split the i.MX8 specific bootloader preparation logic outside of the common post-image into a separate script called as a post-image script.

Best regards,

Thomas
--
Thomas Petazzoni, CTO, Bootlin (formerly Free Electrons) Embedded Linux and Kernel engineering
https://bootlin.com

This series provides initial support for NXP iMX8 (imx8mq-evk).

This series of patches are tested against NXP imx8mq-evk. It only handles basic support for iIMX8M, currently no support for GPU, VPU and other HW features. This series makes it possible to build a minimal system for imx8mq-evk.

Changes since v5
================
- Switch to letting post-image.sh generate imx-boot-imx8mqevk-sd.bin instead of letting imx-mkimage do that.

Changes since v4
================
- Build kernel, u-boot, atf from tag rel_imx_4.9.51_8mq_beta_patch.
- Add board readme.txt.
- Update imx-mkimage package to use host mkimage

Changes since v3
================
- Use Buildroot toolchain instead of external Linareo toolchain.
- Changed title on patch number 2

Changes since v2
================
- Build kernel, u-boot, atf based on tags instead of branches.
- Make sure to assign BR2_PACKAGE_FREESCALE_IMX_PLATFORM to IMX8M.
- Remove kernel defconfig from board folder and use arch default in kernel tree.
- Split third commit into two commits.

Changes since v1
================
- Removed new imx-atf package, use boot/arm-trusted-firmware.
- Move imx-mkimage from package/freescale-imx to package.
- Add upstream URL for imx-mkimage in Config.in.host.
- Fix install command syntax for imx-mkimage package.
- Use Makefile in imx-mkimage to build imx-boot-imx8mqevk-sd.bin instead of doing that in the post-script.sh
- Split third commit into two, one for firmware-img package and one for iMX8 board.

Erik Larsson (4):
imx-mkimage: new package
freescale-imx: add config option for i.MX8M
firmware-imx: add support for iMX8M firmware configs/freescale_imx8mqevk: new defconfig

 .../common/imx/genimage.cfg.template_imx8          | 40 +++++++++++
 board/freescale/common/imx/post-image.sh           | 28 +++++++-
 board/freescale/imx8mqevk/readme.txt               | 83 ++++++++++++++++++++++
 configs/freescale_imx8mqevk_defconfig              | 41 +++++++++++
 package/Config.in.host                             |  1 +
 package/freescale-imx/Config.in                    |  5 +-
 package/freescale-imx/firmware-imx/Config.in       |  2 +-
 package/freescale-imx/firmware-imx/firmware-imx.mk | 13 ++++
 ...-for-overriding-bl32-and-bl33-not-only-bl.patch | 58 +++++++++++++++
 .../0002-add-ldflags-to-link-step.patch            | 26 +++++++
 package/imx-mkimage/Config.in.host                 |  7 ++
 package/imx-mkimage/imx-mkimage.hash               |  3 +
 package/imx-mkimage/imx-mkimage.mk                 | 24 +++++++
 13 files changed, 326 insertions(+), 5 deletions(-)
 create mode 100644 board/freescale/common/imx/genimage.cfg.template_imx8
 create mode 100644 board/freescale/imx8mqevk/readme.txt
 create mode 100644 configs/freescale_imx8mqevk_defconfig
 create mode 100644 package/imx-mkimage/0001-add-support-for-overriding-bl32-and-bl33-not-only-bl.patch
 create mode 100644 package/imx-mkimage/0002-add-ldflags-to-link-step.patch
 create mode 100644 package/imx-mkimage/Config.in.host
 create mode 100644 package/imx-mkimage/imx-mkimage.hash
 create mode 100644 package/imx-mkimage/imx-mkimage.mk

--
2.7.4