fsck.ext3 different?

Stephen C. Tweedie sct at redhat.com
Tue May 3 12:12:39 UTC 2005


Hi,

On Wed, 2005-04-20 at 09:59, Mary Ellen Foster wrote:

> You can mount partitions without running fsck, though -- just put a
> "0" in the last field in /etc/fstab and it'll mount without checking.
> This has worked in my experience, but I don't claim to understand all
> of the issues and maybe this is just papering over some looming
> problem ... but as long as you boot FC4 occasionally, the partition
> will still get checked, right?

Not really.

With a journaling filesystem, "fsck" actually performs two completely
separate functions.  One is the normal recovery that occurs to make the
filesystem consistent after an uncontrolled reboot; the other is a full
consistency check which scans the entire filesystem and corrects any
problems seen.

The recovery bit is very quick, which is why reboots are faster with
journaling filesystems.  And the kernel can do a recovery itself, so if
you don't fsck the filesystem at boot time, it doesn't matter.

But when a full consistency check is required --- either because the
"time until next check" counters have expired (see man tune2fs), or
because the kernel previously encountered a problem on the filesystem
and marked it as containing an error, then you need to have a fsck that
understands all of the current features on that filesystem.

So if you periodically boot to FC4 from an FC2 machine, all that happens
is that if there's no error, nothing gets checked (ie. no improvement);
and if there _is_ an error, then most of your boots are still into FC2,
so at that point the machine won't boot.

You can either keep a newer rescue disk around to do the forced fsck if
you ever get errors on the fs, or simply disable the features you don't
want (with tune2fs), and do a forced fsck to clean up.

--Stephen





More information about the test mailing list