2016-10-17 14:45 GMT+02:00 Marek Vasut marex@denx.de:
On 10/17/2016 02:28 PM, Max Krummenacher wrote:
Hi Marek
#if defined(CONFIG_SPL_MMC_SUPPORT) #define CONFIG_SYS_MMCSD_RAW_MODE_U_BOOT_SECTOR 138 /* offset 69KB */ +#define CONFIG_SPL_PAD_TO 0x11000 /* offset 68KB */
Does this mess up boards which can boot from both SD and other boot media (NAND, SPI NOR, PNOR...) ?
Good point. CONFIG_SPL_PAD_TO is used to create the combined SPL/U-Boot binary. While I have not found any use of this by any config which includes imx6_spl.h it might hinder future boards. So it's probably best to define CONFIG_SPL_PAD_TO in the individual board configs and skip this patch.
So how did you come up with this patch ? What was the thought process?
I'm preparing patches for submission of new boards, Colibri iMX6, Apalis iMX6. They rely on this patchset https://www.mail-archive.com/u-boot@lists.denx.de/msg227944.html, so the board patches are not yet ready for submission.
OT: Can you make those boards load u-boot.img from extfs partition instead ?
I probably could but did not try.
Also it would change the way we use U-Boot across our module family and I'm worried that we more often need to recover over USB due to a corrupted filesystem. On top of that I made some measurements on loading the kernel from an ext2 instead of an fat fs and have seen differences bigger than 100ms. So parsing into a filesystem twice would probably add even more overhead.
Max
I forward ported from a downstream 2015.04 U-Boot where 'make u-boot-with-spl.imx' used an offset of 68KB between SPL and U-Boot binary matching the SPL code which later loads U-Boot. With the current master it used an offset of 64KB resulting in an SPL not loading U-Boot. Probably the 4KB stem from now counting SPL_PAD_TO from the start of the IVT/DCD region and earlier counting from the start of SPL.
I see, thanks for clarifying.
Max
-- Best regards, Marek Vasut