Hi Francesco, On Wed, Oct 14, 2026 at 10:52 Francesco Dolcini wrote: > Which features? Something you are going to send to mainline U-Boot or not?
CMD_NANDBCB, which the Vybrid nandbcb series adds for mainline, and on our products the USB Ethernet gadget for flashing over the USB cable and the CDC ACM gadget for a console on the same cable. The gadgets are existing U-Boot features, so only config. colibri_vf_defconfig with these three is 498200 bytes, 6.7 KB over the limit. The Vybrid gives us no hardware root of trust, so for the EU Cyber Resilience Act we want signed FIT images, a write list for the environment and a password on the autoboot. These add 34 KB to colibri_vf_defconfig, 21 KB over the limit. > Did you consider moving from a monolitic U-Boot to SPL+U-Boot proper > instead? Would that help? It would, but U-Boot has no SPL support for the Vybrid: ARCH_VF610 does not select SUPPORT_SPL and vf610_nfc has no SPL loader. SPL would also mean two images in NAND and a second update path. The DCD gets U-Boot into DDR with the code that is there, one image and the same NAND layout, as the i.MX6 boards without SPL do. > Are you going to send also the config change to enable LTO? Not planned; the 26 KB it gains do not cover the features above, and with the DCD it is not needed. > Is there any known downside or limitation from this change? - The DCD cannot branch on the module, so the DDR clock is PLL2 PFD2 on both; the VF50 took it from the system clock before, at the same 396 MHz. Tested on both modules. - A wrong DCD stops the ROM before U-Boot prints anything; the serial downloader still recovers the module. imx_usb_loader from Boundary Devices hands the DCD to the ROM and works; the older toradex/imx_loader fork emulates the DCD with plain writes only and refuses its set, clear and check entries. - A module still on the WinCE eBoot works: eBoot 1.9 writes the image with "flashloader" and the ROM boots it from NAND, tested on both modules. > Last, whatever is going to be the outcome of this decision, please > always have the "why" of the change in the commit message. The "what" can > be normally see from the code. Will do. We have a VF50 and a VF61 on an Iris carrier on the bench, with serial console, USB recovery and Linux on both. If there is anything else you would like to see tested before you decide, I am happy to run it. Regards, Mehmet
