Hi Petr, On 2026-10-10T14:24:46, Petr Hodina via B4 Relay <[email protected]> wrote: > rockchip: LZMA-compress the FDT image in the FIT based on FIT_FDT_COMP > > Adds FIT_FDT_COMP alongside the existing FIT_UBOOT_COMP gated on > CONFIG_SPL_LZMA only. Applies it to the @fdt-SEQ template's compression > property.
Please start with a paragraph giving the motivation (and imperative mood while you are there): which board needs this, and how much space does it save in the FIT? Compressing the FDT also adds decompression time in SPL, so before/after numbers for size and boot time would help. Please mention the target board in the cover letter too. Please also explain why only LZMA is used. I assume it is because patch 1 only implements 'lzma' in binman. If so, say so, since a reader will compare this with FIT_UBOOT_COMP just above, which prefers gzip. > > Signed-off-by: Petr Hodina <[email protected]> > > arch/arm/dts/rockchip-u-boot.dtsi | 8 +++++++- > 1 file changed, 7 insertions(+), 1 deletion(-) > diff --git a/arch/arm/dts/rockchip-u-boot.dtsi > b/arch/arm/dts/rockchip-u-boot.dtsi > @@ -19,6 +19,12 @@ > +#if defined(CONFIG_SPL_LZMA) > +#define FIT_FDT_COMP "lzma" > +#else > +#define FIT_FDT_COMP "none" > +#endif Just to check: as far as I can tell, no in-tree Rockchip defconfig enables CONFIG_SPL_LZMA, so this has no effect on any current board. That is fine, but please mention it in the commit message so people know it is opt-in. This also ties FDT compression to CONFIG_SPL_LZMA. Someone who enables LZMA in SPL for another reason, e.g. a compressed kernel in Falcon mode, will find their FDT compressed without asking for it. If both CONFIG_SPL_GZIP and CONFIG_SPL_LZMA are enabled, U-Boot is gzip-compressed and the FDT is LZMA-compressed. That works, but it may not be what people expect. I wonder if a separate option would be clearer, or just a comment here saying that FDT compression follows CONFIG_SPL_LZMA and why. What do you think? BTW I checked spl_load_fit_image(): the hash is verified on the compressed data before decompression, and mkimage hashes what binman stores, so signed and hashed FITs should be fine. Did you boot-test this with CONFIG_SPL_FIT_SIGNATURE enabled? It would be worth mentioning in the commit message. Regards, Simon
