Switching back from nouveau doesn't work - Was: Proprietary nvidia driver for different kernels
 

Hi :)

I tried to switch back to nv for the kernel-rt, but it failed. GDM
doesn't start anymore.

[root@archlinux spinymouse]# pacman -Rss xf86-video-nouveau
Targets (1): xf86-video-nouveau-1.0.2-1
[root@archlinux spinymouse]# rm /etc/X11/xorg.conf
[root@archlinux spinymouse]# cp /etc/X11/xorg.conf.nv /etc/X11/xorg.conf
[root@archlinux spinymouse]# pacman -Rdds nouveau-dri libgl
Targets (3): libdrm-nouveau1-2.4.33-1 libgl-8.0.4-3
nouveau-dri-8.0.4-2
[root@archlinux spinymouse]# shutdown -r now



GDM won't start anymore, I booted another linux and continued with:



spinymouse@precise:~$ sudo sed -i 's/options nouveau modeset=1/#options
nouveau modeset=1/' /mnt/archlinux/etc/modprobe.d/
sed: can't read /mnt/archlinux/etc/modprobe.d/modprobe.conf: No such
file or directory

spinymouse@precise:~$ ls /mnt/archlinux/etc/modprobe.d/modprobe.conf
ls: cannot access /mnt/archlinux/etc/modprobe.d/modprobe.conf: No such
file or directory

spinymouse@precise:~$ grep nouveau /mnt/archlinux/etc/mkinitcpio.conf
spinymouse@precise:~$



Useless since there's nothing to do. I didn't switch to systemd.



spinymouse@precise:~$ grep (EE /mnt/archlinux/var/log/Xorg.0.log
(WW) warning, (EE) error, (NI) not implemented, (??) unknown.
[ 17.176] (EE) Failed to load module "glx" (module does not exist, 0)
[ 17.178] (EE) Failed to load module "v4l" (module does not exist, 0)
[ 17.247] (EE) NV: The PCI device 0x10de01d3 (GeForce 7300 SE) at
01@00:00:0 has a kernel module claiming it.
[ 17.247] (EE) NV: This driver cannot operate until it has been
unloaded.
[ 17.248] (EE) No devices detected.



Any hints are welcome.

Regards,
Ralf

PS: I did jitter tests using the nouveau driver, but forgot to take care
about the CPU frequency scaling. I don't know if it was set up to
performance, so I need to do the tests again. IOW first I'll install
nouveau again, but I also want to be able to switch back to nv.

This is an older test:
http://www.digipedia.pl/usenet/thread/11091/63/

No Xruns! As long as peak is > 1ms, it's still > 0.2ms, everything
should be ok for my taste,
so even while the current tests are worse compared with the old tests,
peak jitter perhaps still is inaudible, even for me and I'm very
sensitive to timing issues, phasing and things like that.

However, it's strange to get Xruns when doing a MIDI latency test, this
seems to be a serious issue.

And now today's test:
[spinymouse@archlinux ~]$ uname -r
3.4.4-rt14-1-rt

[spinymouse@archlinux ~]$ hdspmixer
Card 0: RME AIO S/N 0x579bcc at 0xfdbf0000, irq 18
Card 1: TerraTec EWX24/96 at 0xbf00, irq 20
Card 2: TerraTec EWX24/96 at 0xbb00, irq 21

[spinymouse@archlinux ~]$ sudo chgrp audio /dev/hpet
[spinymouse@archlinux ~]$ sudo chmod g+rw /dev/hpet
[spinymouse@archlinux ~]$ sudo modprobe snd-hrtimer

[spinymouse@archlinux ~]$ /etc/rc.d/rtirq status | grep n
281 FF 85 - 125 0.0 S irq/18-snd_hdsp
283 FF 80 - 120 0.0 S irq/20-snd_ice1
304 FF 79 - 119 0.0 S irq/21-snd_ice1
330 FF 50 - 90 0.0 S irq/18-nouveau

[spinymouse@archlinux ~]$ jackd --sync -Xalsarawmidi -dalsa -r48000
-p256
jackdmp 1.9.8
JackALSARawMidiDriver::Attach - input port registered (alias='system:1-1
HDSPMx579bcc MIDI 1 1 in').
JackALSARawMidiDriver::Attach - input port registered (alias='system:2-1
TerraTec EWX24/96 MIDI 1 in').
JackALSARawMidiDriver::Attach - output port registered
(alias='system:1-1 HDSPMx579bcc MIDI 1 1 out').
JackALSARawMidiDriver::Attach - output port registered
(alias='system:2-1 TerraTec EWX24/96 MIDI 1 out').

[spinymouse@archlinux ~]$ jack_midi_latency_test -m 3 -s 8192 -t 5
"system:1-1 HDSPMx579bcc MIDI 1 1 out" "system:1-1 HDSPMx579bcc MIDI 1 1
in"
Waiting for connections ...
Waiting for test completion ...

Reported out-port latency: 5.33-5.33 ms (256-256 frames)
Reported in-port latency: 5.33-5.33 ms (256-256 frames)
Average latency: 11.72 ms (562.26 frames)
Lowest latency: 11.67 ms (560 frames)
Highest latency: 11.84 ms (568 frames)
Peak MIDI jitter: 0.17 ms (8 frames)
Average MIDI jitter: 0.03 ms (0.23 frames)

Jitter Plot:
0.0 - 0.1 ms: 8185
0.1 - 0.2 ms: 2
> 10 ms: 5

Latency Plot:
11.6 - 11.7 ms: 1068
11.7 - 11.8 ms: 7109
11.8 - 11.9 ms: 10
> 21.6 ms: 5

Messages sent: 8192
Messages received: 8192
Xruns: 5

[spinymouse@archlinux ~]$ jack_midi_latency_test -m 3 -s 8192 -t 5
"system:2-1 TerraTec EWX24/96 MIDI 1 out" "system:2-1 TerraTec EWX24/96
MIDI 1 in"
Waiting for connections ...
Waiting for test completion ...

Reported out-port latency: 5.33-5.33 ms (256-256 frames)
Reported in-port latency: 5.33-5.33 ms (256-256 frames)
Average latency: 11.72 ms (562.71 frames)
Lowest latency: 11.69 ms (561 frames)
Highest latency: 11.81 ms (567 frames)
Peak MIDI jitter: 0.12 ms (6 frames)
Average MIDI jitter: 0.01 ms (0.09 frames)

Jitter Plot:
0.0 - 0.1 ms: 8192

Latency Plot:
11.6 - 11.7 ms: 1
11.7 - 11.8 ms: 8187
11.8 - 11.9 ms: 4

Messages sent: 8192
Messages received: 8192

Switching back from nouveau doesn't work - Was: Proprietary nvidia driver for different kernels
 

On Tue, 2012-09-18 at 16:58 +0200, I made some typos:
> No Xruns! As long as peak is > 1ms, it's still > 0.2ms, everything
> should be ok for my taste

*chuckle* The ">" should be "<"

Switching back from nouveau doesn't work - Was: Proprietary nvidia driver for different kernels
 

Hi :)

I set up CPU frequency scaling as needed and run the Jack MIDI latency
tests again. I suspect the nouveau driver still is broken [1]. There are
still Xruns and there's an absurd high peak.

Now I need to get my nvidia/nv set up back, but this script
https://wiki.archlinux.org/index.php/NVIDIA#Switching_between_nvidia_and_nouveau_driver s is useless, because of

[spinymouse@archlinux ~]$ cat /etc/modprobe.d/modprobe.conf
cat: /etc/modprobe.d/modprobe.conf: No such file or directory
[spinymouse@archlinux ~]$ grep MODULES /etc/mkinitcpio.conf
# MODULES
# MODULES="piix ide_disk reiserfs"
MODULES=""
# 'filesystems' is _required_ unless you specify your fs modules in
MODULES
## This setup specifies all modules in the MODULES setting above.

Since the nv driver still is installed, I also tried to blacklist
nouveau, but starting GDM always failed.

I prefer not to restore Arch from a backup. There must be a way to get
rid of nouveau and get nv and nvidia working again.

At the moment I'm stuck.

Regards,
Ralf

[1]
[spinymouse@archlinux ~]$ cpupower frequency-info | grep The gov
The governor "ondemand" may decide which speed to use

[spinymouse@archlinux ~]$ sudo cpupower frequency-set -g performance

[spinymouse@archlinux ~]$ cpupower frequency-info | grep The gov
The governor "performance" may decide which speed to use

[spinymouse@archlinux ~]$ uname -r
3.4.4-rt14-1-rt

[spinymouse@archlinux ~]$ hdspmixer
Card 0: RME AIO S/N 0x579bcc at 0xfdbf0000, irq 18
Card 1: TerraTec EWX24/96 at 0xbf00, irq 20
Card 2: TerraTec EWX24/96 at 0xbb00, irq 21

[spinymouse@archlinux ~]$ sudo chgrp audio /dev/hpet
[spinymouse@archlinux ~]$ sudo chmod g+rw /dev/hpet
[spinymouse@archlinux ~]$ sudo modprobe snd-hrtimer

[spinymouse@archlinux ~]$ /etc/rc.d/rtirq status
PID CLS RTPRIO NI PRI %CPU STAT COMMAND
59 FF 90 - 130 0.0 S irq/8-rtc0
300 FF 85 - 125 0.0 S irq/18-snd_hdsp
335 FF 80 - 120 0.0 S irq/20-snd_ice1
338 FF 79 - 119 0.0 S irq/21-snd_ice1
116 FF 70 - 110 0.0 S irq/19-ehci_hcd
118 FF 70 - 110 0.5 S irq/16-ohci_hcd
121 FF 69 - 109 0.0 S irq/17-ohci_hcd
125 FF 68 - 108 0.0 S irq/17-ohci_hcd
57 FF 65 - 105 0.0 S irq/1-i8042
20 FF 50 - 90 0.0 S irq/9-acpi
105 FF 50 - 90 0.1 S irq/22-ahci
123 FF 50 - 90 0.0 S irq/18-ohci_hcd
127 FF 50 - 90 0.0 S irq/18-ohci_hcd
130 FF 50 - 90 0.0 S irq/14-pata_ati
131 FF 50 - 90 0.0 S irq/15-pata_ati
267 FF 50 - 90 0.0 S irq/7-parport0
293 FF 50 - 90 0.0 S irq/22-firewire
330 FF 50 - 90 0.0 S irq/18-nouveau
588 FF 50 - 90 0.0 S irq/43-eth0
3 FF 1 - 41 0.0 S ksoftirqd/0
11 FF 1 - 41 0.0 S ksoftirqd/1

[spinymouse@archlinux ~]$ jackd --sync -Xalsarawmidi -dalsa -r48000
-p256
jackdmp 1.9.8

[spinymouse@archlinux ~]$ jack_midi_latency_test -m 3 -s 8192 -t 5
"system:1-1 HDSPMx579bcc MIDI 1 1 out" "system:1-1 HDSPMx579bcc MIDI 1 1
in"

Reported out-port latency: 5.33-5.33 ms (256-256 frames)
Reported in-port latency: 5.33-5.33 ms (256-256 frames)
Average latency: 11.69 ms (560.52 frames)
Lowest latency: 11.65 ms (559 frames)
Highest latency: 11.76 ms (564 frames)
Peak MIDI jitter: 0.10 ms (5 frames)
Average MIDI jitter: 0.02 ms (0.04 frames)

Jitter Plot:
0.0 - 0.1 ms: 8189
> 10 ms: 3

Latency Plot:
11.6 - 11.7 ms: 7906
11.7 - 11.8 ms: 283
> 21.6 ms: 3

Messages sent: 8192
Messages received: 8192
Xruns: 3

[spinymouse@archlinux ~]$ jack_midi_latency_test -m 3 -s 8192 -t 5
"system:2-1 TerraTec EWX24/96 MIDI 1 out" "system:2-1 TerraTec EWX24/96
MIDI 1 in"
Waiting for connections ...
Waiting for test completion ...

Reported out-port latency: 5.33-5.33 ms (256-256 frames)
Reported in-port latency: 5.33-5.33 ms (256-256 frames)
Average latency: 11.69 ms (561.05 frames)
Lowest latency: 11.40 ms (547 frames)
Highest latency: 11.76 ms (564 frames)
Peak MIDI jitter: 0.35 ms (17 frames)
Average MIDI jitter: 0.00 ms (0.04 frames)

Jitter Plot:
0.0 - 0.1 ms: 8191
0.2 - 0.3 ms: 1

Latency Plot:
11.4 - 11.5 ms: 1
11.6 - 11.7 ms: 7704
11.7 - 11.8 ms: 487

Messages sent: 8192
Messages received: 8192

Switching back from nouveau doesn't work - Was: Proprietary nvidia driver for different kernels
 

On Tue, 18 Sep 2012 22:41:32 +0200
Ralf Mardorf <ralf.mardorf@alice-dsl.net> wrote:

> I set up CPU frequency scaling as needed and run the Jack MIDI latency
> tests again. I suspect the nouveau driver still is broken [1]. There
> are still Xruns and there's an absurd high peak.

At this moment in time I'll skip the rest of your msg since I don't
know all that much about midi nor about switching back and forth
between nouveau and nvidia.

As the maintainer and author of the current linux-rt package both on
aur and in the archaudio-production repo, I'd be interested in getting
the midi performance tested and approved though

One thing did catch my attention though. you have your rme audio card
in a slot that shares irq with the nvidia and with several usb
devices. Imo that is really asking for trouble and you need to at the
very least get your soundcard to use another interrupt than the nvidia
card...

--

Joakim

Switching back from nouveau doesn't work - Was: Proprietary nvidia driver for different kernels
 

On Sat, 2012-09-22 at 18:00 +0200, Joakim Hernberg wrote:
> One thing did catch my attention though. you have your rme audio card
> in a slot that shares irq with the nvidia and with several usb
> devices. Imo that is really asking for trouble and you need to at the
> very least get your soundcard to use another interrupt than the nvidia
> card...

Yes, but I don't had this issue in the past. I can't get another IRQ for
the sound card. FWIW, the latency test did work with the TerraTec too,
there was some "Voodoo" when it didn't work using the nv driver.

However, I try to get rid of the NVIDIA card. I will remove it and check
if Linux nowadays is able to enable 3D support for the on-board ATI
Radeon X1250-based graphics.

Regards,
Ralf

Switching back from nouveau doesn't work - Was: Proprietary nvidia driver for different kernels
 

> However, I try to get rid of the NVIDIA card. I will remove it and check
> if Linux nowadays is able to enable 3D support for the on-board ATI
> Radeon X1250-based graphics.

Just thought I'd give a heads up that a major overhaul of nouveau
apparently including knowledge gained from the last few years is planned
for release with kernel 3.7.

http://www.h-online.com/open/features/Kernel-Log-Major-overhaul-of-Nouveau-1664869.html

--
__________________________________________________ _____________________

'Write programs that do one thing and do it well. Write programs to work
together. Write programs to handle text streams, because that is a
universal interface'

(Doug McIlroy)
__________________________________________________ _____________________