Dear all

I'm using a AT91SAM9260-EK board. I managed to replace the jffs2 rootfs to a yaffs2 rootfs ( of course, I also changed the boot args in u-boot). To get the yaffs2 image on the device, I booted into ramdisk, mounted the yaffs rootfs and untared the rootfs on the yaffs partition.

This works flawless - but on production environment, this procedure is way to slow for image flashing. Is there known working way to get the image on the device using yaffs2-utils and u-boot? Or is there another _fast_ way to get the root fs on the device (except flashing the chips while manufacturing...)

Kind regards,

Mmame

What's to stop you from TFTP'ing the finished file system directly in-place within the NAND flash? I don't understand the necessity to unpack it from a ramdisk, as this suggests you've already delivered a semi-working image or linux, to the board.

You could use a donor-image created from your "golden" system, if you are unable to generate the final image mechanically on the host PC.

In production, you could also use a gold image of the entire active NAND array (non-blank blocks) from your golden system, which would include AT91BootStrap, U-Boot, Environment, Kernel and File System(s). This could be written en-masse to the NAND, delivered via TFTP, or other means.

If you had thought about this during board design, you could have a header to plug in an external production boot programming board. Such a board could either have a dataflash device on board which RomBoot would find first, or a NOR device setting the BMS pin to boot strap your own ROM code. The board could either contain the entire gold image, or be able to pull it rapidly from another server, or whatever.

Absent this thought, you could use SAM-BA, the DLLs, or your own SAM-BA type application to configure SDRAM, load a custom applet into it, and then pull your gold image via serial/usb/ethernet or whatever to the production device.

In production you want to automate as much of this as possible, to minimize human interaction, and censored ups.

Now I'm going to presume that the 9260-EK board is not what you are going to production with, but I'll note that ATMEL did have an 8MB DataFlash device that plugged into the MMC socket that could be used to boot strap the board as I recall.

See here :
http://www.atmel.com/dyn/resources/prod ... oc3257.pdf
Not a standard MMC device, but using the SPI connection, and form-factor.