Peter Gork wrote:
> Dear JFS group,
>
> I have a question on the implementation of JFS as root file system on
Linux. Consider the case JFS is used
> as root file system and is not unmounted cleanly.
> JFS does replay its log in an user program, not in the kernel. How does
it apply the log before mounting a JFS
> root file system?
>
> It seems to me that JFS will mount a potentially
> inconsistent / JFS read-only first and then only later do the log replay
in the file system checker with logredo. > Is that correct?

One problem that could happen is that the file system is in such a bad
state that it can't
find /sbin/fsck.jfs, thus logredo can't be run. Are you seeing this type of
problem?

Couple solutions:
1) have fsck.jfs be loaded by initrd.img.
2) bring the logredo functionality into the file system.

> Doesn't that cause problems
> with the kernel image of the fs getting inconsistent
> with the result of the log replay?
>

Steve

_______________________________________________
Jfs-discussion mailing list
[EMAIL PROTECTED]
http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jfs-discussion

Reply via email to