mounted external ext4 causes "high" kworker cpu

[Jeff Moyer]

Chris Murphy <lists at colorremedies.com> writes:

> On Apr 27, 2012, at 9:53 AM, Jeff Moyer wrote:
>
>> Chris Murphy <lists at colorremedies.com> writes:
>> 
>>> Normally top reports CPU line, sy at 0.4% when idle. If I format an
>>> external Firewire disk as btrfs and mount it, it remains at 0.4%. If I
>>> reformat as XFS and mount it, again top reports sy at 0.4%. However,
>>> if I reformat as ext4 and mount it, sy runs at 3.5%. These two
>>> processes are now at the top of top's results:
>>> 
>>> kworker/1:2
>>> kworker/0:4
>>> 
>>> Each uses on average 1.9% CPU. The light on the external drive flashes
>>> 4x per second. There are no processes using the disk at all while this
>>> is going on. If I umount it, the pulsing stops. If I remount it, the
>>> pulsing resumes as does the slightly higher CPU consumption.
>>> 
>>> This doesn't happen with the same hardware mounted XFS or btrfs or
>>> HFS+.
>>> 
>>> Odd?
>> 
>> Sounds like lazy itable init.  Try mkfs -t ext4 -E lazy_itable_init=0.
>
> This is happening on internal disks, targeted for Fedora
> installation. I can hear it once install is complete, still booted
> from DVD media. Is this a bug?

So long as the file system is mounted and the itable initialization is
not complete, I would expect the initialization to continue.  Does that
answer your question?

Cheers,
Jeff