Eep. Child "ncp.cfg" (ino #344) of dir ino #88 doesn't exist!

Eep. Child "ncp.cfg" (ino #344) of dir ino #88 doesn't exist!

[jm1]

JFFS2 on a powerPC embedded system. 64M Intel Strataflash. Latest CVS patched
into Hardhat Linux (2.4.3 based)

We are using the board for development and test. It had an uptime of about a
month and was working fine - it was running CVS code from 6 to 8 weeks ago.

Today we rebooted it with the latest MTD CVS code (as of 5/2/02) and got:

Eep. Child "ncp.cfg" (ino #344) of dir ino #88 doesn't exist!

in the syslog. I am unable to delete the bad entry from the dir inode:

# perl -e "unlink 'ncp.cfg';"
# ls -la
ls: ./ncp.cfg: Input/output error
drwxrwxr-x 2 root root 0 May 3 14:10 .
drwxr-xr-x 8 root root 0 May 3 14:57 ..

Rebooting doesn't seem to help - the error is reported each time.

Other than this error the JFFS2 filesystem is working fine.

Obviously I can clear this up by erasing the whole flash, but is there anything
I should look at first?

Thanks,

Tom Angert

Re: Eep. Child "ncp.cfg" (ino #344) of dir ino #88 doesn't exist!

[David Woodhouse]

jm1@angert.net said:
> Today we rebooted it with the latest MTD CVS code (as of 5/2/02) and
> got:

> Eep. Child "ncp.cfg" (ino #344) of dir ino #88 doesn't exist!

> in the syslog.

OK, that looks like filesystem corruption of some kind - you have a 
directory entry for inode #344, but the inode itself doesn't exist.

Does it printk any other clues as to why that happened? Nodes with a bad 
CRC, perhaps?

> I am unable to delete the bad entry from the dir inode:

> # perl -e "unlink 'ncp.cfg';" 
> # ls -la 
> ls: ./ncp.cfg: Input/output

That's very odd. Can you print the return code (and errno) of the unlink() 
syscall? That really ought to work.

Can you compile with CONFIG_JFFS2_FS_DEBUG=1, set /proc/sys/kernel/printk 
to 9 and log the kernel's output during mount and unlink attempts?

I assume the old kernel is showing the same behaviour if you reboot back to 
it?



--
dwmw2