Problem mounting JFFS2
Hello
I'm trying to get a JFFS2 filesystem on a 64MB NAND flash device working on an ARM based system (running Linux-2.4.21 and the latest MTD CVS code).
I prepare the JFFS2 image on my host system like this
temp> mkfs.jffs2 --version
mkfs.jffs2: revision 1.38
temp> mkfs.jffs2 -d targetfs -o targetfs.img -e 16 -l -q -v -n
and burn it into the NAND device on the target. If i later try to mount it under Linux with "mount -t jffs2 /dev/mtdblock/1 /mnt", i get the following output:
jffs2: Erase block size too small (16KiB). Using virtual blocks size (32KiB) insteadjffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block at 00000000
OOB at 00000000 was 33 f3 c0 ff f0 ff ff ff ff ff ff ff ff ff ff ff jffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block at 000080
OOB at 00008000 was 03 f0 00 ff f0 ff ff ff ff ff ff ff ff ff ff ff [deleted rest of cleanmarker check]
Ino #14 is absent but node not REF_PRISTINE. Reading. ARGH. About to write node to 0x00004000 on flash, but there are data already there:0x00004000: 85 19 02 e0 20 03 00 00 10 25 f5 2e 0e 00 00 00 Write of 1576 bytes at 0x00004000 failed. returned -5, retlen 0 Not marking the space at 0x00004000 as dirty because the flash driver returned retlen zero
Error writing new dnode: -5
Any idea what is going wrong?
TIA
Dave
Quoted 10 lines · David Müller (ELSOFT AG)
Hello
I'm trying to get a JFFS2 filesystem on a 64MB NAND flash device working on an ARM based system (running Linux-2.4.21 and the latest MTD CVS code).
I prepare the JFFS2 image on my host system like this
temp> mkfs.jffs2 --version
mkfs.jffs2: revision 1.38
Quoted 6 lines
temp> mkfs.jffs2 -d targetfs -o targetfs.img -e 16 -l -q -v -n
and burn it into the NAND device on the target. If i later try to mount it under Linux with "mount -t jffs2
/dev/mtdblock/1/mnt", i get the following output:
How do you burn it into the NAND device ? Are you updating the OOB area ? read http://linux-mtd.infradead.org/tech/nand.html
use nandwrite from the util directory, if you can do it from a running system
Quoted 2 lines
jffs2: Erase block size too small (16KiB). Using virtual blocks size (32KiB) instead
That's due to memory consumption. If we have 16KiB eraseblock size we need too much memory to keep track of all eraseblocks. So we use virtual 32KiB blocks instead.
Quoted 7 lines
jffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block at 00000000
OOB at 00000000 was 33 f3 c0 ff f0 ff ff ff ff ff ff ff ff ff ff ffjffs2_check_nand_cleanmarker(): Cleanmarker node not detected in block at 000080
OOB at 00008000 was 03 f0 00 ff f0 ff ff ff ff ff ff ff ff ff ff ff [deleted rest of cleanmarker check]
See above
Signature · 5 lines
Thomas ________________________________________________________________________ linutronix - competence in embedded & realtime linux http://www.linutronix.de mail: [email protected]
Hello
Quoted 15 lines · Thomas Gleixner
Quoted 13 lines · David Müller (ELSOFT AG)
Hello
I'm trying to get a JFFS2 filesystem on a 64MB NAND flash device working on an ARM based system (running Linux-2.4.21 and the latest MTD CVS code).
I prepare the JFFS2 image on my host system like this
temp> mkfs.jffs2 --version
mkfs.jffs2: revision 1.38temp> mkfs.jffs2 -d targetfs -o targetfs.img -e 16 -l -q -v -n
Quoted 10 lines
Quoted 3 lines
and burn it into the NAND device on the target. If i later try to mount it under Linux with "mount -t jffs2
/dev/mtdblock/1/mnt", i get the following output:How do you burn it into the NAND device ? Are you updating the OOB area ? read http://linux-mtd.infradead.org/tech/nand.html
use nandwrite from the util directory, if you can do it from a running system
I'm using U-Boot's http://sourceforge.net/projects/u-boot "nand" command to erase and write the NAND flash.
If i use the nandwrite command
# nandwrite --version nandwrite 1.2
# nandwrite --jffs2 /dev/mtd/1 /tmp/targetfs.img
and mount it, the error is gone but the mounted filesystem is empty (no files, not directories).
If i write the image using U-Boot's command, then there is the reported error but the mounted filesystem is populated with directories and some files (see below) although some files are missing or corrupt.
# ll /mnt/ total 0 drwxr-xr-x 2 0 0 0 Apr 23 2003 bin drwxr-xr-x 2 0 0 0 Apr 2 2003 dev drwxr-xr-x 2 0 0 0 Apr 2 2003 dev-state drwxr-xr-x 2 0 0 0 Apr 2 2003 etc drwxr-xr-x 2 0 0 0 Apr 2 2003 home drwxr-xr-x 2 0 0 0 Apr 2 2003 lib drwxr-xr-x 2 0 0 0 Apr 2 2003 proc drwxr-xr-x 2 0 0 0 Apr 2 2003 root drwxr-xr-x 2 0 0 0 Apr 2 2003 sbin drwxr-xr-x 2 0 0 0 Apr 2 2003 tmp drwxr-xr-x 2 0 0 0 Apr 2 2003 usr drwxr-xr-x 2 0 0 0 Apr 2 2003 var # ll /mnt/bin/ total 391 -rwxr-xr-x 1 0 0 400168 Apr 23 2003 armdemo
Quoted 9 lines
Quoted 2 lines
jffs2: Erase block size too small (16KiB). Using virtual blocks size (32KiB) instead
That's due to memory consumption. If we have 16KiB eraseblock size we need too much memory to keep track of all eraseblocks. So we use virtual 32KiB blocks instead.
Yes, i know. This was the reason why i had to upgrade to the latest version of the JFFS2 code, as the version i used before was unable to handle it.
What really bothers me is this error:
Ino #14 is absent but node not REF_PRISTINE. Reading. ARGH. About to write node to 0x00004000 on flash, but there are data already there:0x00004000: 85 19 02 e0 20 03 00 00 10 25 f5 2e 0e 00 00 00 Write of 1576 bytes at 0x00004000 failed. returned -5, retlen 0 Not marking the space at 0x00004000 as dirty because the flash driver returned retlen zero
Error writing new dnode: -5
Dave