[Buildroot] Dataflash rootfs problems (still)

[Ulf Samuelsson]

ons 2007-10-24 klockan 18:54 +0200 skrev Jorge S.:
> 
> 
> 2007/10/24, Ulf Samuelsson <ulf at atmel.com>:
>         ons 2007-10-24 klockan 18:20 +0200 skrev Jorge S.:
>         > Hello all:
>         >
>         > I've sucessfully built  several root filesystems  for my
>         custom board
>         > (AT91RM9200 + 8MiB DataFlash),  but  now  i'm facing
>         a  strange 
>         > problem.
>         >
>         > I'm using a JFFS2 filesystem on an MTD partition into the
>         DataFlash,
>         > my rootfs boots OK, but ONLY for the first time. See steps:
>         >
>         > 1) Create a rootfs JFFS2 image using buildroot -> OK 
>         > 2) Erase entire flash partition: I boot from an initrd
>         containing
>         > mtd-tools for this purpouse. -> OK
>         > 3) Dump the rootfs image "rootfs.jffs2" into the Dataflash
>         (i use
>         > loadb from u-boot here...) -> OK 
>         > 4) Boot the system... -> OK
>         > 5) Login and issue a simple reboot, no other
>         commands...nothing
>         > else.... -> OK
>         >
>         
>         If you run the SPI faster than 5 Mbps, you are basically
>         screwed.
>         See Errata #13 in the datasheet.
>         
>         You will need to add an external AND gate if you want to run
>         faster.
>         
>         * AND output to the dataflash chip select,
>         * one input to the NPCS0
>         * one input to a GPIO
>         
>         The driver needs to be adapted for this.
> 
> So, there's a flash corruption on the 1st boot?, 
> 
> I'm a little confused now...why u-boot looks like writing OK then?
> I've sucessfully uploaded hundreds of kernels and initrds images to
> the dataflash. 
> 
> During the 1st boot, Dropbear creates and "writes" the generated keys,
> maybe is this *write* operation the one screwing up the dataflash?
> 
> And last question, which one is the best way to solve this by
> software? I suppose i can "slow down" the SPI speed somewhere in the
> kernel or board config at u-boot isn't it? 
> 
> Thanks for helping Ulf.

It is one potential problem, which needs to be addressed,
in U-Boot, linux and dataflashboot.bin.
There may be other problems, which I am not aware of,  but
it is best to fix this to start with.


-- 
Best Regards,
Ulf Samuelsson