linux/meta-freescale.git/recipes-bsp/imx-atf/imx-atf_2.14.bb, branch master

imx-atf: Build out of tree and from a clean directory

2026-10-28T19:26:25+00:00

imx-atf built in the source tree, into ${S}/build, and do_compile never cleaned it. The TF-A Makefile does not track build options such as SPD or LOG_LEVEL, so after a PACKAGECONFIG change make finds every object up to date and relinks nothing. do_unpack does not depend on PACKAGECONFIG, so it does not rerun either, and do_deploy then ships the bl31.bin from the previous configuration. On imx8mq-evk this showed up after switching from the NXP BSP to the mainline BSP, which drops the optee PACKAGECONFIG. The deployed BL31 was still the SPD=opteed build. Its imx8mq setup sets the BL32 entry to BL32_BASE unconditionally, and the mainline boot container carries no OP-TEE, so the board hung right after the "BL31: Built" banner. The stale binary was reproduced byte for byte by building with SPD=opteed and then without it in the same build directory. Build into ${WORKDIR}/build through BUILD_BASE, clean it before every compile, and deploy from there. The Makefile stays in ${S}, so do_compile runs make with -C ${S}; the default do_compile would skip make, as ${B} has no Makefile. Tested on imx8mq-evk with the mainline BSP by building imx-atf with the optee PACKAGECONFIG and then without it. Each build recompiles everything, the optee bl31 carries the opteed service and the plain one does not. With u-boot-fslc 2026.10, core-image-weston then boots on an imx8mq-evk board through BL31, U-Boot and Linux to Weston. Signed-off-by: Rodrigo M. Duarte

imx-atf: Update to lf-6.18.37_2.1.0

2026-10-26T13:28:56+00:00

Update to the NXP LF6.18.37_2.1.0 release (9 commits), which adds the initial i.MX 937 support and several SCMI and TRDC fixes. Drop two patches that are no longer needed: - 0001-scmi-imx9-Remove-unused-tmp-pointer.patch no longer applies: upstream e5d744b06 ("LF-18583 fix(scmi): fix chunked transfer in scmi_per_lpm_mode_set") reworked the function, and it now uses tmp, so the GCC 16 unused-variable error is gone. - 0001-imx93-trdc-Fix-header-guard.patch was no longer listed in SRC_URI, and the fix is already upstream. Tested by building imx-atf for imx8mq-evk, imx8mp-lpddr4-evk, imx93-11x11-lpddr4x-evk, imx943-19x19-lpddr4-evk and imx95-19x19-verdin. Signed-off-by: Rodrigo M. Duarte

imx-atf: add generic ATF_SYS_PWR_FULL_CTRL passthrough

2026-10-01T14:25:10+00:00

BL31's SYS_PWR_FULL_CTRL (plat/imx/imx9/{imx94,imx95}/platform.mk) selects whether SCMI shutdown/cold-reset requests ask for a full SoC/board power-down or reset. Add ATF_SYS_PWR_FULL_CTRL as a machine-overridable variable, mirroring the existing ATF_BOOT_UART_BASE/IMX_BOOT_UART_BASE pattern in this same recipe. Defaults to "0", matching BL31's own default. Signed-off-by: Francesco Dolcini

imx-atf: Add CVE_PRODUCT

2026-09-17T13:19:09+00:00

imx-atf sets no CVE_PRODUCT, so cve-check falls back to "imx-atf". That name is in no NVD record, so the recipe is always reported with zero CVEs. imx-atf is NXP's fork of TF-A. Use the same CVE_PRODUCT as meta-arm's trusted-firmware-a.inc. Signed-off-by: Hiago De Franco

imx-atf: Update to 2.14

2026-09-14T12:01:44+00:00

Move the downstream NXP TF-A from the lf_v2.12 branch to lf_v2.14, which is the ARM Trusted Firmware matching the NXP lf-6.18.20_2.0.0 release. Refresh 0001-scmi-imx9-Remove-unused-tmp-pointer.patch: the SCMI vendor driver moved from drivers/arm/css/scmi/vendor/scmi_imx9.c to drivers/nxp/scmi/vendor/scmi_imx9.c in 2.14, so the patch no longer applied. The unused-but-set tmp pointer in scmi_per_lpm_mode_set() is still present, so the fix is still required to build with GCC 16 and -Werror. Signed-off-by: Stefan Eichenberger