Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes empty_blocks , can't read superblock
Marco Braga
marco.braga at gmail.com
Tue Feb 17 08:53:20 EST 2009
Hi,
I have exactly the same problem. The problem is in the organization of
the free bytes in the spare area and the way flash_eraseall expects
them to be. I've not had the time to check the issue in every detail,
but this is what I remember from my brief check. In our board ECC
bytes are put near the middle of the 16 byte spare area (we use 3 byte
ecc). The problem is that flash_eraseall do not agree with JFFS2 in
the way to use the remaining 13 bytes. It believes that it can put an
8 byte cleanmark (since 13 bytes are unused), but there is not enough
contiguous space to but an 8 byte cleanmarker. Notice that
flash_eraseall does not support a split cleanmark so it writes a wrong
mark.
When you mount the device erased with flash_eraseall, then JFFS2
complains about this.
Try this (your exact numbers will vary):
# flash_eraseall -j /dev/mtd7
Erasing 16 Kibyte @ 7fc000 -- 99 % complete. Cleanmarker written at 7fc000.
# nanddump -p -s 0x7fc000 -l 1 /dev/mtd7
ECC failed: 0
ECC corrected: 0
Number of bad blocks: 0
Number of bbt blocks: 0
Block size 16384, page size 512, OOB size 16
Dumping data starting at 0x007fc000 and ending at 0x007fc001...
0x007fc000: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc010: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc020: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc030: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc040: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc050: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc060: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc070: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc080: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc090: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc0a0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc0b0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc0c0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc0d0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc0e0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc0f0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc100: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc110: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc120: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc130: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc140: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc150: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc160: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc170: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc180: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc190: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc1a0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc1b0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc1c0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc1d0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc1e0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
0x007fc1f0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
OOB Data: 85 19 03 20 08 ff ff ff ff ff ff ff ff ff ff ff
Notice that cleanmarker is only 5 bytes long... Because bytes 6,7,8
are used by our ECC code. The problem is that flash_eraseall expects
to write a linear (contiguous) marker.
This is the relevant code:
/* Check for autoplacement */
if (oobinfo.useecc == MTD_NANDECC_AUTOPLACE) {
/* Get the position of the free bytes */
if (!oobinfo.oobfree[0][1]) {
fprintf (stderr, " Eeep.
Autoplacement selected and no empty space in oob\n");
exit(1);
}
clmpos = oobinfo.oobfree[0][0];
clmlen = oobinfo.oobfree[0][1];
if (clmlen > 8)
clmlen = 8;
} else {
/* Legacy mode */
switch (meminfo.oobsize) {
case 8:
clmpos = 6;
clmlen = 2;
break;
case 16:
clmpos = 8;
clmlen = 8;
break;
case 64:
clmpos = 16;
clmlen = 8;
break;
}
}
Both methods are basically wrong.
2009/2/10 <santhoshunnikrishnan at tataelxsi.co.in>:
>
> Hi all,
>
> I am having a problem in using with flash_erasell -j and mount commands
> together.
>
> I erased one NAND partition using the command "flash_eraseall -j /dev/mtd1"
>
> The erase was successfully completed and the clean up marker were written
> (as per the prints , i am saying this)
>
> Then i tried to mount the erased partition using the command "mount -t jffs2
> /dev/mtdblock1 /mnt"
>
> but i got one error message saying that
>
> "Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes
> empty_blocks 8, bad_blocks 0, c->nr_blocks 48
> mount: /dev/mtdblock1: can't read superblock"
>
> I tried to erase the flash using the command "flash_eraseall /dev/mtd1"
>
> and then i tried to mount the erased partition using the command "mount -t
> jffs2 /dev/mtdblock1 /mnt"
>
> This time it successfully mounted and i am able to write contents to the
> partition.
>
> I want to know why i am not able to mount after erasing the flash using the
> -j option. I know that -j is to write the cleanup marker and the MTD site
> recommands erasing the flash with this "-j"
>
> Can anyone help me what the problem is ?
>
>
> Regards,
> Santhosh Unnikrishnan
>
>
> The information contained in this electronic message and any attachments to this message are intended for the exclusive use of the addressee(s) and may contain proprietary, confidential or privileged information. If you are not the intended recipient, you should not disseminate, distribute or copy this e-mail. Please notify the sender immediately and destroy all copies of this message and any attachments contained in it.
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
>
More information about the linux-mtd
mailing list