Question about yocto, imx28evk, MfgTool and NAND

persmitt · ‎08-18-2014

Hi

I am having trouble with the spam filter so I will split this question into several replies on the same thread to figure out why the spam filter hates me ;-)

I am getting started with an embedded Linux project using Yocto and the i.MX28 evk. So far things have gone pretty well but I have run into a hurdle when trying to get Linux to boot from the NAND flash instead of the SD card.

Before I go more into the questions I will just summarize quickly of what I have done to get a general idea of where I stand.

For starters I followed Yocto Training - HOME to get started with the Yocto concept. I took the generated image, dd it to an SD card and then started Linux on the eval board without any problems. I also played around by adding an ftp-server, sshd and apache just to learn a bit more how to add packages.

persmitt · ‎08-21-2014

After my latest post I had a u-boot and a kernel flashed to NAND. I was however mistaken about the load address. Fortunately someone with more knowledge had prepared U-boot with default parameters.

=> printenv loadaddr

loadaddr=0x42000000

I have no clue why it is at 0x42000000 and not at 0x40008000 that I got from uboot-mkimage but the U-boot preloaded value was correct in my case. Anyone with input on this would be appreciated.

Once the kernel has the correct load address it will get much further during startup, all the way to mounting the root file system. For the file system it seemed that ubifs was the modern way to go so thats what I did.

Flashing a ubifs file system on a nand can be done in two ways. You can, for one, flash it with a generic NAND flasher. This requires a ubi image which is created by running ubinize on an existing ubifs image.

The other alternative is to flash a ubifs image through U-Boot, which is what I decided to do. The ubifs image is created by running mkubifs but this is handled automatically by bitbake if you have set ubifs as a target in conf/machine/imx28evk.conf.

IMAGE_FSTYPES ?= "tar.bz2 ext3 uboot.mxsboot-sdcard sdcard ubifs"

However building the ubifs image requires certain parameters set in the same file (conf/machine/imx28evk.conf)

MKUBIFS_ARGS = "--min-io-size 2048 --leb-size 126976 --max-leb-cnt 864"

These values can be found by running ubiattach -m 1 /dev/ubi_ctrl (requires packages mtd-utils and mtd-utils-ubifs) in Linux on your target system. The easiest way to do this is to boot on an SD card, see Task #4 - Deploy and test

However depending on your bootargs you might not have access to /dev/mtdX. If the mtd device is missing, reboot and enter U-Boot and add gpmi (Can not see the NAND device /dev/mtd0 or /dev/mtd1 for micron MT29F1G08ABADA) to the mmcargs.

setenv mmcargs setenv bootargs gpmi ${mtdparts} console=${console_mainline},${baudrate} root=${mmcroot}

run mmcboot

Your /dev/mtdX should be available now. The NAND partitions does not work like normal file system partitions, instead they are entirely logical and setup with an address and a size. In U-Boot the NAND partitions can be setup neatly with the command mtdparts and for many imx boards that can translate just nicely to a booted Linux, but not the imx28. The other boards have an external file where the NAND partitions can be configured but the imx28 has the partition hardcoded in the driver. It can be found (grep for mtd_partition) and modified but then it must be maintained as a kernel patch every time there is an upgrade. Instead I opted for going with the default partition the imx28 nand driver offers: /dev/mtd0 20MiB, /dev/mtd1 has the rest. Verify the numbers using mtdinfo command. My strategy is to put bootloader, environment and kernel in /dev/mtd0 and in /dev/mtd1 I put filesystems by using multiple ubifs volumes in the same partition.

So analyzing the output from ubiattach -m 1 /dev/ubi_ctrl

UBI: attaching mtd1 to ubi0

UBI: physical eraseblock size: 131072 bytes (128 KiB)

UBI: logical eraseblock size: 126976 bytes

UBI: smallest flash I/O unit: 2048

UBI: VID header offset: 2048 (aligned 2048)

UBI: data offset: 4096

UBI: attached mtd1 to ubi0

UBI: MTD device name: "gpmi-nfc-general-use"

UBI: MTD device size: 108 MiB

UBI: number of good PEBs: 864

UBI: number of bad PEBs: 0

UBI: max. allowed volumes: 128

UBI: wear-leveling threshold: 4096

UBI: number of internal volumes: 1

UBI: number of user volumes: 0

UBI: available PEBs: 852

UBI: total number of reserved PEBs: 12

UBI: number of PEBs reserved for bad PEB handling: 8

UBI: max/mean erase counter: 1/1

UBI: image sequence number: 0

UBI: background thread "ubi_bgt0d" started, PID 397

UBI device number 0, total 864 LEBs (109707264 bytes, 104.6 MiB), available 852 LEBs (108183552 bytes, 103.2 MiB), LEB size 126976 by)

Use the values to set the MKUBIFS_ARGS to get a ubifs image that will work for your NAND flash.

Now since we are reconsidering the partitioning lets recap where to put stuff and start from the point where we only have u-boot on the nand.

First lets build the partition table using mtdparts default and then we delete the incorrectly placed file system.

mtdparts default

mtdparts del filesystem

Then we place the file system after the first 20MiB. I use a 128MiB NAND here.

mtdparts add nand0 0x06C00000@0x1400000 filesystem

The partition table should look like this now

=> mtdparts

device nand0 <gpmi-nand>, # parts = 7

#: name size offset mask_flags

0: bootloader 0x00300000 0x00000000 1

1: environment 0x00080000 0x00300000 0

2: redundant-environment0x00080000 0x00380000 0

3: kernel 0x00400000 0x00400000 0

4: fdt 0x00080000 0x00800000 0

5: ramdisk 0x00800000 0x00880000 0

6: filesystem 0x06c00000 0x01400000 0

Lets save the partition table to the environment.

saveenv

Now let us first flash the kernel where it is supposed to be. As usual we download it from a USB thumb since I haven't gotten to setup TFTP yet.

usb start

fatls usb 0:1

nand erase.part kernel

fatload usb 0:1 ${loadaddr} uimage-imx28evk.bin

nand write ${loadaddr} kernel

Now the kernel is in place, lets setup the kernel boot arguments. U-boot comes with a default environment nandargs which is almost entirely correct for us.

nandargs=setenv bootargs console=${console_mainline},${baudrate} rootfstype=ubifs ubi.mtd=6 root=ubi0_0 ${mtdparts}

The part that is incorrect for us is ubi.mtd=6 as we are planning to have the filesystem on mtd1. So we take the setenv part and runs it manually. It is also important to add the gpmi argument so the nand driver will load.

setenv bootargs gpmi console=${console_mainline},${baudrate} rootfstype=ubifs ubi.mtd=1 root=ubi0_0 ${mtdparts}

setenv bootcmd 'nboot kernel; bootm'

saveenv

We can now try to boot and should get as far as trying to mount the root file system.

So now it is time to put in our ubifs image into nand. First we load the ubifs image into memory.

usb start

fatls usb 0:1

fatload usb 0:1 ${loadaddr} core-image-minimal-imx28evk.ubifs

Then we erase the filesystem partition.

nand erase.part filesystem

We set the ubi subsystem to work with the filesystem partition

ubi part filesystem

We create a volume. Since we don't set type or size it becomes dynamic and covers the entire size.

ubi create rootfs

Write the loaded image to the newly created volume. If we had created multiple volumes we chose which volume to receive the image. Useful for having multiple filesystems/partitions.

ubi write ${loadaddr} rootfs ${filesize}

Finally we mount the file system locally to verify that it looks ok.

ubifsmount ubi0:rootfs

ubifsls

Now we can reboot and this time Linux should start properly. It did for me and I hope it will work for you too!

View solution in original post

sonuabraham · ‎06-01-2017

Hi Per Smitt,

Can you let us know how the following crash in the rootfs was solved . I am getting similar error when trying to boot an Angstrong image on imx28evk board.

undefined instruction

pc : [<40008010>] lr : [<47f38440>]

sp : 47b32d70 ip : 00000000 fp : 00000000

r10: 00000000 r9 : 47b32f38 r8 : 47b417dc

r7 : 00000000 r6 : 40008000 r5 : 47f9f724 r4 : 00000000

r3 : 47b32fe0 r2 : 40000100 r1 : 000009e3 r0 : 47b32fe0

Flags: nZCv IRQs off FIQs off Mode SVC_32

Resetting CPU ...

Thanks in advance,

Sonu Abraham

lzh_kc · ‎10-30-2015

Hi persmitt ，

You said you had "played around by adding an ftp-server, sshd and apache just to learn a bit more how to add packages",

I'm now trying to add apcahe,mysql and php into my image, but don't know how.

Here are what i have tryed:

I do bitbake php first, and then add "php" and "php-cli" (which inspired by an answer in question Re: Adding php package to fsl-image-machine-test ) into the CORE_IMAGE_EXTRA_INSTALL varible in local.conf, then do "bitbake fsl-image-qt5". I run the image on my imx6 board and happiely find that php works fine.

But when i want to add apache and mysql, that's not work, I find a .bb file for apache: "source/meta-openembedded/meta-webserver/recipes-httpd/apache2/apache2_2.4.10.bb", but I don't know how to use it.

As to mysql, I do as adding php, and I get only a .so file in a deep chlid tree in /usr/lib on the imx6 board.

So how do I get wrong , how do you add apache? Please help me, many thanks!

@persmitt

persmitt · ‎10-30-2015

Just add apache2 to the same variable. If you find a recipe (in this case apache2_2.4.10.bb) simply truncate the name at the underscore and add it to your CORE_IMAGE_EXTRA_INSTALL. That should add apache to your build.

persmitt · ‎08-21-2014