mmcqd process is causing hickups / small freezes of the system

[GoogleCodeExporter]

Hi,

I've been porting gnuboy to dingux and i noticed that the dingux system
would occasionally freeze a fraction of a second and then go on. 

It's noticeable in a lot of games and emulators.

I was intrigued by the problem so i took some time to check up on it.

I started an emulator / game and ran top from within a telnet session to
see the process info. Whenever the hickups occur the mmcqd process goes up
in the list to use a huge amount of cpu power, causing games and emulators
to free / slowdown for a fraction of second. IT's really irritating.

Steps to reproduce :
take a cpu heavy application let it run or play a game a bit. Also run top
from within a telnet session, now keep looking at top while you play and
when it freezes you'll notice the mmcqd process kicking in.

Steps to (possibly) resolve the issue ?
i did some googling around and found this page describing the issue for
another embedded system : 
http://www.nabble.com/mmcqd-uses-a-large-amount-of-cpu-td23768078.html

the suggested solution there was to mount the partition in async mode. So i
mounted the rootfs file, copied it to a new directory and added the async
option in the inittab file and made a new image using genromfs, the image
boots fine but i have no way in telling if it's mounted in async mode
neither do i know if it will fix the issue at hand. 

please look into this 
thanks 
joyrider

Original issue reported on code.google.com by joyrider...@hotmail.com on 5 Oct 2009 at 12:11

[GoogleCodeExporter]

i read on some other page that lowering the process default scheduling priority 
might
help as well :

renice 20 `pidof mmcqd`

but i still noticed it kicking in sometimes. isn't there a way to disable this 
or
stop this from happening ?

Original comment by joyrider...@hotmail.com on 5 Oct 2009 at 1:15

[GoogleCodeExporter]

Original comment by igga...@gmail.com on 5 Oct 2009 at 3:26

• Changed state: Accepted

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

mounting /boot with noatime and nodiratime helps as a lot as well but it still 
occurs
although it has less of an impact now. I'm using it in combination with the 
renice 20
`pidof mmcqd` command and it doesn't seem to happen so often anymore (but still 
does
from time to time).

these parameters i'm using now for the mount -o remount command in inittab :
rw,nosuid,nodev,noatime,nodiratime

Also (offtopic) aren't the filesystem corruptions related to user software ? 
let me
explain the filesystem is probably mounted async so it's up to the software / 
user to
call flush to write changes back to the filesystem and to the sd card. If this 
isn't
done it can cause corruption (the gp2x had this problem as well). it could 
still be
something else also but it's just a thought.

Original comment by joyrider...@hotmail.com on 5 Oct 2009 at 7:21