7 Oct
2007
7 Oct
'07
12:23 p.m.
On Friday 05 October 2007 18:14, Nigel Henry wrote:
On Thursday 04 October 2007 21:29, Roman Muñoz wrote:
As promised, an update on realtime on Debian Lenny. That is after having
installed from the musix repo, kernel 2.6.21-rt4, and 2.6.21.5-rt18.
First I had to update util-linux from the unstable repo, as the version on
Lenny was the Etch one, and still thought that schedutils was available. The
Sid (unstable) version installed ok, and chrt is available now.
Then added the lines below to /etc/security/limits.conf.
@audio - rtprio 99
@audio - nice -10
@audio - memlock 4000000
Now the fun started with installing rtirq (the version from the musix repo).
It installed, but refused to start. I was confused, as both scripts, the one
in /etc/init.d, and /etc/sysconfig seemed to be the same, so I booted up
Fedora 7, and checked the 2 rtirq scripts there, and sure enough the one
in /etc/sysconfig was a very short one, compared to the start script
in /etc/init.d.
Downloaded the rtirq tarball from the blog that Roman posted.
http://www.rncbc.org/jack/rtirq-20070101.tar.gz
Unpacked it, deleted the duplicated startup script in Lenny's /etc/sysconfig,
pasted the correct config script in /etc/sysconfig, renaming it from
rtirq.conf to rtirq.
Rebooted, and ran /etc/init.d/rtirq status, and got the following output.
/etc/init.d/rtirq status
PID CLS RTPRIO NI PRI %CPU STAT COMMAND
184 FF 90 - 130 0.0 S< IRQ-8 rtc
666 FF 85 - 125 0.0 S< IRQ-9 ohci1394, bttv0, Bt87x audio
1770 FF 84 - 124 0.1 S< IRQ-10 EMU10K1
631 FF 80 - 120 0.0 S< IRQ-11 uhci_hcd:usb1, uhci_hcd:usb2,
eth0
288 FF 75 - 115 0.0 S< IRQ-1 i8042
287 FF 74 - 114 0.0 S< IRQ-12 i8042
4 FF 50 - 90 0.0 S softirq-high/0
5 FF 50 - 90 0.0 S softirq-timer/0
6 FF 50 - 90 0.0 S softirq-net-tx/
7 FF 50 - 90 0.0 S softirq-net-rx/
8 FF 50 - 90 0.0 S softirq-block/0
9 FF 50 - 90 0.0 S softirq-tasklet
10 FF 50 - 90 0.0 S softirq-sched/0
11 FF 50 - 90 0.0 S softirq-hrtimer
12 FF 50 - 90 0.0 S softirq-rcu/0
207 FF 50 - 90 3.2 S< IRQ-14 ide0
210 FF 50 - 90 0.0 S< IRQ-15 ide1
785 FF 50 - 90 0.0 S< IRQ-6 floppy
1581 FF 50 - 90 0.0 S< IRQ-7 parport0
3078 FF 50 - 90 0.0 S< IRQ-3
Jackd has a prio of 70, and just with jackd started, only one xrun up to now
after 58mins of 0.118 msecs duration.
Job done. Realtime on Etch, and Lenny. The above makes it look dead easy, but
there was a lot of frustration along the way, and it took longer than it
appears.
Many thanks to all those who offered help on this list, and the debian-user
list.
One question regarding the rtirq status output above.
IRQ-9 has a slightly higher rtprio than that for the soundcard on IRQ-10.
ohci1394 is tied in with the firewire port on the audigy2 soundblaster, but I
can probably stop that being loaded, as I don't use firewire. bttv0 is the TV
card, and BT87x audio is for the audio capture part of the TVcard.
What would I need to do to swap these around, so that IRQ-9 has a prio of 84,
and IRQ-10, a prio of 85?
It's probably a bit of an academic question, as your unlikely to be doing
serious audio work while watching TV.
Nigel.
Hi Nigel,
Is rtirq working four you? If yes, I think that your soundcard IRQ
should be at prio 85 at startup... and it was not.
See this:
http://mggmail.blogspot.com/2007/07/re-rtirq-fixed.html
Well I've got realtime working ok on Debian Etch, and Rick Wright's one
liner in rc.local is changing the prio to 90 for the soundcard on bootup.
But, rtirq is something different. The version installed is the same as on
the blog above (20070101), and the 2 files are installed, one
in /etc/init.d/rtirq, and the other in /etc/sysconfig/rtirq. I believe
the /etc/sysconfig directory was created when installing rtirq, as I don't
rmember it being there before. I say this because with Fedora the sysconfig
directory exists, and there is a bunch of files in it, and I remember
looking in Debian for this directory, and couldn't find it.
I also ran sysv-rc-conf as root on Etch. The rtirq entry is there, but
runlevels 2,3,4,and5 are unchecked, so something isn't quite right here. I
see also on the blog above that rtirq should be in /etc/default, but in the
case of Etch it isn't. As I say. Realtime on Etch is working ok, but I'll
go through the same procedure on Debian Lenny (testing), and try and fix
the rtirq problem without using Rick Wright's one liner, although it's much
appreciated, and works like clockwork.
I'll post back after I've played around with realtime on Lenny a bit.
Best regards,
Roman
Thanks for all the replies to this interesting subject.
Nigel.