failed to boot jffs2-filesystem from nand flash in TMS320DM8168

hi,

I do not use the SPI Flash, So I have a try with a kernel disable the SPI flash support. Following is the boot log, I think maybe the problem is always the ecc, because I sam a display in the boot log as "jffs2_scan_eraseblock(): Node at 0x00e029c8 {0x1985, 0xe041, 0x00000031) has invalid CRC 0x4282d91d (calculated 0x4ddad371)"

5758.spi.txt

When you created rd-jffs2.bin file, which root fs you have pass to it?

# mkfs.jffs2 -lqn –e 128 -r target -o /tftpboot/rd-jffs2.bin

What you pass for "-r target"? What is the "target"?

I use the  rootfs decompressed from "arago-base-tisdk-image-dm816x-evm.tar.gz" and In order to make the rootfs be smaller, I remove the "arago-base-tisdk-image-dm816x-evm.tar.gz" under the decompressed directory.

Fire Lord said:

Following is the boot log, I think maybe the problem is always the ecc, because I sam a display in the boot log as "jffs2_scan_eraseblock(): Node at 0x00e029c8 {0x1985, 0xe041, 0x00000031) has invalid CRC 0x4282d91d (calculated 0x4ddad371)"

You are flashing the default linux kernel, just remove SPI flash support from it, nothing else changed? Do you roll back from 8-bit ECC to 1-bit ECC? If no, can you try leaving the u-boot with the default BCH8 (do not switch to 1-bit ECC when booting the board).
 Can you provide me the exact steps when flashing the kernel?

Can you provide me the exact steps when flashing the rd-jffs2.bin file?

BR
Pavel

You are flashing the default linux kernel, just remove SPI flash support from it, nothing else changed? 

BR
Pavel

1.remove SPI support; 2.change host usb to gadget usb; 3.change the ecc_opt,like following pictures:

/arch/arm/ mach-omap2/board-flash.c

/drivers/mtd/nand/omap2.c:omap_nand_probe()

1.flashing u-boot:

mmc rescan 0

mw.b 0x81000000 0xFF 0x260000

fatload mmc 0 81000000 u-boot.noxip.bin

nand erase 0x0 0x260000

nandecc hw 2

nand write.i 0x81000000 0x0 0x260000

nandecc hw 0

2.flashing kernel:

mw.b 0x81000000 0xFF 0x440000

fatload mmc 0 81000000 uImage

nand erase 0x00280000 0x00440000

nand write 0x81000000 0x00280000 0x00351000

3.flashing fs:

mw.b 0x81000000 0xFF 0x00ED0000

fatload mmc 0 81000000 rd-jffs2.bin

nand erase clean 0x006C0000 0x0C820000

nand write 0x81000000 0x006C0000 0x00ED0000

4.boot

I have tried to boot using BCH0 and BCH8 and SW. All failed.

Fire Lord,

Can we first try the below setup:

- never switch to 1-bit ECC (BCH0), stay with BCH8 from end to end
- keep the default kernel, just remove SPI Flash and USB host, keep it in BCH8 (default)

How to flash uImage from u-boot:
TI8168_EVM#mw.b 0x81000000 0xFF 0x440000
TI8168_EVM#tftp 0x81000000 uImage
TI8168_EVM#nand erase 0x00280000 0x00440000
TI8168_EVM#nand write 0x81000000 0x00280000 <uImage size, in example 0x280000>

How to flash rd-jffs2.bin from u-boot:
TI8168_EVM#mw.b 0x81000000 0xFF 0x0C820000
TI8168_EVM#tftp 0x81000000 rd-jffs2.bin
TI8168_EVM#nand erase clean 0x006C0000 0x0C820000
TI8168_EVM#nand write 0x81000000 0x006C0000 <rd-jffs2.bin szie, in example 0xE01000>

How to set up bootargs and boot:

TI8168_EVM#setenv bootargs console=ttyO2,115200n8 root=/dev/mtdblock3 rw rootfstype=jffs2 mem=100M earlyprintk
TI8168_EVM#nand read 0x81000000 0x00280000 <uImage size, in example 0x280000>
TI8168_EVM#bootm 0x81000000

Let me know if this works for you.

BR
Pavel

hi,

can you tell about how to enable the tftp function?

I can only get the infomation like below. But I do not know how to set on my computer.

hi,

hello,pavel,

I can not get across the tftp setting. So can you teach me about that?

I use the SD card "fatload" to replace "tftp", what's the differences between these two ways?

I have tested with SD, here is the boot log:

5141.full_test.txt

After about two minutes, the login interface will come out. I can login but with nothing under the directory of "root". For normal login of SD boot, there will be a 'dm816x-evm' in the directory of 'root'.

whit's more, after a few minutes, it will come out the warning of 

Read of newly-erased block at 0x0b040000 failed: -74. Putting on bad_list
Read of newly-erased block at 0x0b040000 failed: -74. Putting on bad_list
JFFS2: marking eraseblock at 0b040000
as bad

Can you explain that?

'

Fire Lord said:

In the boot log, we can see "WARNING: at drivers/dsp/syslink/omap_notify/notify.c:776 notify_send_event+0xd4/0x26c()". Does it mean there is something wrong with DSP?

Actually keep the boot args as we already discussed:

console=ttyO2,115200n8 root=/dev/mtdblock3 rw rootfstype=jffs2 mem=100M earlyprintk

This error occur as the udev is trying to insert the vpss module in the kernel, but there are several steps that should be done prior to insert this vpss module. To workaround this problem:

1. Delete the vpss.ko module from the arago fs: /lib/modules/2.6.37/kernel/drivers/video/ti81xx/vpss/vpss.ko

2. Delete the sii9022a.ko module from the arago fs: /lib/modules/2.6.37/kernel/drivers/video/ti81xx/sii9022a/sii9022a.ko

3. Create new rd-jffs2.bin file based on the new arago fs (with delete modules from it)

4. Flash the new rd-jffs2.bin file in NAND

Thus you will boot up properly, but not have vpss.ko and sii9022a.ko auto inserted by udev in the kernel. If you need these modules inserted in your kernel, you can do this manually after the board boot up, follow the below wiki pages how to do:

Regards,
Pavel

hi,

Glad to tell you I have saw hope of success.

First try:

"""Can you try to add in the bootrgs 'notifyk.vpssm3_sva=0xBF900000'? If does not work, try with 'notifyk.vpssm3_sva=0xA0000000'""""

I tried this for both BCH0 and BCH8 which failed. Here is the boot log:

8270.txt

6087.txt

0523.txt

7331.txt

Second try: I made a new fs under your guide ( delete the two kos)---rd1-jffs2.bin

Similarly, I tested BCH0 and BCH8. When use BCH0, coming the problem as """mtd->read(0x800 bytes from 0xa02000) returned ECC error"""", When use BCH8, coming the problem as """"Read of newly-erased block at 0x09de0000 failed: -74. Putting on bad_list     Read of newly-erased block at 0x09de0000 failed: -74. Putting on bad_list    JFFS2: marking eraseblock at 09de0000     as bad"""". And when use BCH8, this problem isn't coming immediately but after about 3 minutes. I have tried to scrub the all nand which do not take effect. Here is the boot log:

3581.txt

6136.txt

And are you sure that BCH8 is not support jffs2 filesystem? If yes, need i test for BCH8 when boot from nand with jffs2?If yes, the most important and right thing I should do is to solve the problem of ""mtd->read(0x800 bytes from 0xa99000) returned ECC error"". Do I get it?

Another question, if I boot from nand successfully, can I see a diretory called "dm8168-evm" under the directory of "/home/root/" which can be seen from SD boot?Because I can not confirm that if I have boot from nand successfully.

Fire Lord said:

Another question, if I boot from nand successfully, can I see a diretory called "dm8168-evm" under the directory of "/home/root/" which can be seen from SD boot?Because I can not confirm that if I have boot from nand successfully.

This is strange. By default, there is no "dm8168-evm" under the directory of "/home/root/" of the default arago fs.

When you boot up from NAND, check about the presence of the vpss.ko and sii9022a.ko. These two modules should miss in NAND, and present in SD.

Regards,
Pavel

Fire Lord said:

Similarly, I tested BCH0 and BCH8. When use BCH0, coming the problem as """mtd->read(0x800 bytes from 0xa02000) returned ECC error"""", When use BCH8, coming the problem as """"Read of newly-erased block at 0x09de0000 failed: -74. Putting on bad_list     Read of newly-erased block at 0x09de0000 failed: -74. Putting on bad_list    JFFS2: marking eraseblock at 09de0000     as bad"""". And when use BCH8, this problem isn't coming immediately but after about 3 minutes. I have tried to scrub the all nand which do not take effect. Here is the boot log:

Fire Lord said:

And are you sure that BCH8 is not support jffs2 filesystem? If yes, need i test for BCH8 when boot from nand with jffs2?If yes, the most important and right thing I should do is to solve the problem of ""mtd->read(0x800 bytes from 0xa99000) returned ECC error"". Do I get it?

Seems to me you have two options here:

1. Move to UBIFS

2. Roll back to BCH0 (1-bit ECC).

See the below e2e thread which explains in details your error and how to resolve.

Regards,
Pavel

Fire Lord said:

What means to "Roll back to BCH0(1-bit ECC)"? Does it mean I should flashing kernel and fs with BCH0 and boot with BCH0? If yes, Do you have any suggestion to solve the problem of "mtd->read(0x800 bytes from 0xaeb000) returned ECC error"?

This one is fully described in the below e2e thread:

e2e.ti.com/.../239237