The first commit in this series is a fix for qcom_parse_memory, which is currently broken on master (since fc37a73e6679). It's an alternative to the patch Caleb proposed a couple of days ago (20250117102734.3725009-2-caleb.connolly@linaro.org). This approach results in a smaller diff, and also makes it possible to do the thing introduced in the second commit.
The second commit introduces the ability to boot with an internal DT that lacks a fully specified /memory node, instead obtaining that information from the FDT supplied by the previous bootloader.
The first commit should ideally land even if the second one needs more rounds before it's ready. I opted to submit both commits together as one series because, as noted earlier, the way the fix is done introduces functionality that the second commit relies on (specifically: passing the pointer of the specific FDT to parse for its /memory node).
I've tested both these commits together on an msm8916 device and an sdm845 device. The msm8916 device booted with an internal DT that relies on /memory in the external FDT to properly function. The sdm845 device booted with an entirely external FDT. So I reckon I hit all the various code paths in both qcom_parse_memory and board_fdt_blob_setup.
Signed-off-by: Sam Day me@samcday.com --- Sam Day (2): mach-snapdragon: pass fdt to qcom_parse_memory mach-snapdragon: support parsing memory info from external FDT
arch/arm/mach-snapdragon/Kconfig | 13 ++++++++ arch/arm/mach-snapdragon/board.c | 68 ++++++++++++++++++++++++++-------------- 2 files changed, 58 insertions(+), 23 deletions(-) --- base-commit: 639cd409987acf173eaffebe7876968b42fd7c32 change-id: 20250120-qcom-parse-memory-updates-96ffe248cdf1
Best regards,