Re: mkfs.ubifs / ubiformat: Bad UBI magic

[Adrian Hunter <adrian.hunter@nokia.com>]

Tobias Wahlström wrote:
> Hi!
> 
> This is my first post at this list - I hope it makes (at least) some
> sense anyway. I have looked through the mailing archive and searched
> using google without finding anything useful on the topic. I hope I
> provided the information you need - if not, just tell me what you are
> missing!
> 
> I'm working on a project where we would like to use ubi and ubifs as
> root filesystem since it seems to be the best choise in many ways. But
> I have a problem which I can't find anything about either the mail
> archive nor using google etc. I currently use linux 2.6.26 with some
> patches. My target is the freescale imx255 pdk board. My build system
> is Ubuntu 9.04 (run within vmware).
> 
> I have cloned the mtd-utils.git (today 2009-09-05) and compiled the
> tools both for my build system and my target (both without xattr
> support). Then I created a ubifs image using mkfs.ubifs on my build
> system and tried to flash it to the nand using ubiformat. The commands
> I used was:
> 
> build> mkfs.ubifs -m 2048 -e 256KiB -c 384 -U --root rootfs -o rootfs.ubifs

Try making the -e option the leb-size not the peb-size i.e. probably 252KiB

> target> flash_eraseall /dev/mtd2
> target> ubiformat /dev/mtd2 -s 2048 -v -f rootfs.ubifs
> 
> The output from ubiformat is:
> 
> ubiformat: mtd2 (nand), size 100663296 bytes (96.0 MiB), 384
> eraseblocks of 262144 bytes (256.0 KiB), min. I/O size 2048 bytes
> libscan: start scanning eraseblocks 0-384
> libscan: scanning eraseblock 0: empty
> [...]
> libscan: scanning eraseblock 383: empty
> libscan: finished, mean EC 0, 0 OK, 0 corrupted, 384 empty, 0 alien, bad 0
> ubiformat: 384 eraseblocks are supposedly empty
> ubiformat: will write 13 eraseblocks
> ubiformat: eraseblock 0: erase, change EC to 0ubiformat: error!: bad
> UBI magic 0x31181006, should be 0x55424923
> ubiformat: error!: bad EC header at eraseblock 0 of "/rootfs.ubifs"
> 
> If I look into the source code I see that UBI_EC_HDR_MAGIC is defined
> twice in the source both in include/mtd/ubi-media.h and seems to be
> used by all tools except for mkfs.ubifs. The command mkfs.ubifs use
> UBIFS_NODE_MAGIC in mkfs.ubifs/ubifs-media.h. The values do not match
> at all. And it also seems like mkfs.ubifs use little endian to encode
> the magic number and the other tools use big endian... Strange!
> 
> /Tobias
> 
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
>