18 Jul
2018
18 Jul
'18
11:15 p.m.
On 18.07.2018 19:58, Ralf Mardorf wrote:
/etc/init.d/rtirq status
PID CLS RTPRIO NI PRI %CPU STAT COMMAND
441 FF 50 - 90 0.0 S irq/29-mei_me
527 FF 50 - 90 0.0 S irq/31-iwlwifi
8 TS - 0 19 0.0 S ksoftirqd/0
18 TS - 0 19 0.0 S ksoftirqd/1
24 TS - 0 19 0.0 S ksoftirqd/2
30 TS - 0 19 0.0 S ksoftirqd/3
36 TS - 0 19 0.0 S ksoftirqd/4
42 TS - 0 19 0.0 S ksoftirqd/5
48 TS - 0 19 0.0 S ksoftirqd/6
54 TS - 0 19 0.0 S ksoftirqd/7
grep RTIRQ_NAME_LIST /etc/default/rtirq
# RTIRQ_NAME_LIST="rtc snd usb i8042" # old
RTIRQ_NAME_LIST="snd usb i8042"
18 Jul
18 Jul
11:45 p.m.
New subject: reducing xruns (System configuration)
I have this note in my rtirq configuration file, it may be useful,
especially the reference to the web page at the end with discussion by
rtirq author:
# Note found on rnbc.org:
# take note that rtirq:
# a. only honors one and only one driver thread per IRQ line,
# so if IRQ share the same line only one can be prioritized
# b. the "snd" particle applies to PCI audio devices;
# "usb" is generic for (all) USB host controllers and applies
# for all USB devices including audio ones; if you want
# specific ports only to get prioritized, then use the respective
# trait (eg. "usb1", "usb2", etc. you'll have to dig which one
# exactly is right, coz it depends on which physical USB host port
# is the device plugged in).
# see http://www.rncbc.org/drupal/node/850
--
Chris Caudle
19 Jul
19 Jul
12:38 a.m.
New subject: reducing xruns (System configuration)
On Wed, 18 Jul 2018 21:15:21 +0200, Max wrote:
On 18.07.2018 19:58, Ralf Mardorf wrote:
That's most likely the culprit or at least the major culprit. Nothing
has got rt priority.
Are you sure that you are running the lowlatency kernel? If not, than
you need the boot option 'threadirqs'.
What's the output of
uname -rm
Before you boot the lowlatency kernel or a generic kernel with
boot option 'threadirqs', make USB the head and remove the keyboard from
this list, IOW change it to RTIRQ_NAME_LIST="usb snd".
/etc/init.d/rtirq status
PID CLS RTPRIO NI PRI %CPU STAT COMMAND
441 FF 50 - 90 0.0 S irq/29-mei_me
527 FF 50 - 90 0.0 S irq/31-iwlwifi
8 TS - 0 19 0.0 S ksoftirqd/0
18 TS - 0 19 0.0 S ksoftirqd/1
24 TS - 0 19 0.0 S ksoftirqd/2
30 TS - 0 19 0.0 S ksoftirqd/3
36 TS - 0 19 0.0 S ksoftirqd/4
42 TS - 0 19 0.0 S ksoftirqd/5
48 TS - 0 19 0.0 S ksoftirqd/6
54 TS - 0 19 0.0 S ksoftirqd/7
grep RTIRQ_NAME_LIST /etc/default/rtirq
# RTIRQ_NAME_LIST="rtc snd usb i8042" # old
RTIRQ_NAME_LIST="snd usb i8042"
4:20 p.m.
New subject: reducing xruns (System configuration)
On 18.07.2018 22:38, Ralf Mardorf wrote:
On Wed, 18 Jul 2018 21:15:21 +0200, Max wrote:
4.17.8-041708-lowlatency x86_64
On 18.07.2018 19:58, Ralf Mardorf wrote:
That's most likely the culprit or at least the major culprit. Nothing
has got rt priority.
Are you sure that you are running the lowlatency kernel? If not, than
you need the boot option 'threadirqs'.
What's the output of
uname -rm /etc/init.d/rtirq status
PID CLS RTPRIO NI PRI %CPU STAT COMMAND
441 FF 50 - 90 0.0 S irq/29-mei_me
527 FF 50 - 90 0.0 S irq/31-iwlwifi
8 TS - 0 19 0.0 S ksoftirqd/0
18 TS - 0 19 0.0 S ksoftirqd/1
24 TS - 0 19 0.0 S ksoftirqd/2
30 TS - 0 19 0.0 S ksoftirqd/3
36 TS - 0 19 0.0 S ksoftirqd/4
42 TS - 0 19 0.0 S ksoftirqd/5
48 TS - 0 19 0.0 S ksoftirqd/6
54 TS - 0 19 0.0 S ksoftirqd/7
grep RTIRQ_NAME_LIST /etc/default/rtirq
# RTIRQ_NAME_LIST="rtc snd usb i8042" # old
RTIRQ_NAME_LIST="snd usb i8042" Before you boot the lowlatency kernel or a generic
kernel with
boot option 'threadirqs', make USB the head and remove the keyboard from
this list, IOW change it to RTIRQ_NAME_LIST="usb snd"
could you elaborate
how?
7 p.m.
New subject: reducing xruns (System configuration)
On Thu, 19 Jul 2018 14:20:29 +0200, Max wrote:
On 18.07.2018 22:38, Ralf Mardorf wrote:
Hi,
yesno!
I can explain how to change the rtirq configuration line from
RTIRQ_NAME_LIST="snd usb i8042"
to
RTIRQ_NAME_LIST="usb snd"
Run
sudo $EDITOR /etc/default/rtirq
and edit this line.
I can't explain why some 3rd party lowlatency kernel doesn't work with
the installed rtirq script. I'm already not a friend of Ubuntu and
Ubuntu flavours, let alone a friend of 3rd party repositories and
derivatives that are not supported by Ubuntu.
I'm using Arch Linux, but to help novices, I've got an Ubuntu install. I
always try to help even users of Ubuntu derivatives, but it's not
always possible, since it's often way to obscure, what is done by 3rd
party maintainers.
Maybe this lowlatency kernel requires usage of the 'threadirqs' boot
option. If so, add it to your boot loaders configuration file. The
bootloader on my machine is syslinux, this is how it does look like for
the syslinux configuration file:
[rocketmouse@archlinux ~]$ grep -A3 '\^threadirqs' /boot/syslinux/syslinux.cfg
MENU LABEL Arch Linux ^threadirqs
LINUX ../vmlinuz-linux
APPEND root=LABEL=archlinux ro threadirqs
INITRD ../intel-ucode.img,../initramfs-linux.img
I didn't test, if the page table isolation stuff has got impact on
real-time performance, maybe you want to disable it, if so add 'nopti':
[rocketmouse@archlinux ~]$ grep -A3 '\^nopti' /boot/syslinux/syslinux.cfg
MENU LABEL Arch Linux threadirqs ^nopti
LINUX ../vmlinuz-linux
APPEND root=LABEL=archlinux ro threadirqs nopti
INITRD ../intel-ucode.img,../initramfs-linux.img
Btw. your output of /etc/init.d/rtirq status look similar to this one:
[rocketmouse@archlinux ~]$ rtirq status
PID CLS RTPRIO NI PRI %CPU STAT COMMAND
158 FF 90 - 130 0.0 S irq/16-ehci_hcd
163 FF 90 - 130 0.0 S irq/24-xhci_hcd
161 FF 89 - 129 0.0 S irq/23-ehci_hcd
309 FF 85 - 125 0.0 S irq/16-snd_hdsp
327 FF 80 - 120 0.0 S irq/16-snd_ice1
38 FF 50 - 90 0.0 S irq/9-acpi
157 FF 50 - 90 0.0 S irq/1-i8042
164 FF 50 - 90 0.0 S irq/25-ahci[000
275 FF 50 - 90 0.0 S irq/5-parport0
284 FF 50 - 90 0.0 S irq/27-mei_me
288 FF 50 - 90 0.0 S irq/8-rtc0
298 FF 50 - 90 0.0 S irq/18-i801_smb
356 FF 50 - 90 0.0 S irq/28-i915
376 FF 50 - 90 0.0 S irq/29-snd_hda_
541 FF 50 - 90 0.0 S irq/26-enp3s0
8 TS - 0 19 0.0 S ksoftirqd/0
21 TS - 0 19 0.0 S ksoftirqd/1
[rocketmouse@archlinux ~]$ uname -rm
4.17.7-1-ARCH x86_64
IOW USB should be on top of the rt priority list.
What's the output of 'systemctl status rtirq'?
[rocketmouse@archlinux ~]$ systemctl status rtirq
● rtirq.service - Realtime IRQ thread system tuning
Loaded: loaded (/usr/lib/systemd/system/rtirq.service; enabled; vendor preset:
disabled)
Active: active (exited) since Wed 2018-07-18 05:31:47 CEST; 1 day 11h ago
Main PID: 864 (code=exited, status=0/SUCCESS)
Tasks: 0 (limit: 4915)
Memory: 0B
CGroup: /system.slice/rtirq.service
Jul 18 05:31:47 archlinux systemd[1]: Starting Realtime IRQ thread system tuning...
Jul 18 05:31:47 archlinux rtirq[864]: Setting IRQ priorities: start [ehci_hcd] irq=16
pid=158 prio=90: OK.
Jul 18 05:31:47 archlinux rtirq[864]: Setting IRQ priorities: start [ehci_hcd] irq=23
pid=161 prio=89: OK.
Jul 18 05:31:47 archlinux rtirq[864]: Setting IRQ priorities: start [xhci_hcd] irq=24
pid=163 prio=90: OK.
Jul 18 05:31:47 archlinux rtirq[864]: Setting IRQ priorities: start [snd_hdsp] irq=16
pid=309 prio=85: OK.
Jul 18 05:31:47 archlinux rtirq[864]: Setting IRQ priorities: start [snd_ice1] irq=16
pid=327 prio=80: OK.
Jul 18 05:31:47 archlinux systemd[1]: Started Realtime IRQ thread system tuning.
Regards,
Ralf
--
pacman -Q linux{,-rt{-cornflower,,-securityink,-pussytoes}}|cut -d\ -f2
4.17.7-1
4.16.18_rt9-1
4.16.15_rt7-1
4.16.12_rt5-1
4.14.34_rt27-1
On Wed, 18 Jul 2018 21:15:21 +0200, Max wrote:
4.17.8-041708-lowlatency x86_64
On 18.07.2018 19:58, Ralf Mardorf wrote:
That's most likely the culprit or at least the major culprit. Nothing
has got rt priority.
Are you sure that you are running the lowlatency kernel? If not, than
you need the boot option 'threadirqs'.
What's the output of
uname -rm /etc/init.d/rtirq status
PID CLS RTPRIO NI PRI %CPU STAT COMMAND
441 FF 50 - 90 0.0 S irq/29-mei_me
527 FF 50 - 90 0.0 S irq/31-iwlwifi
8 TS - 0 19 0.0 S ksoftirqd/0
18 TS - 0 19 0.0 S ksoftirqd/1
24 TS - 0 19 0.0 S ksoftirqd/2
30 TS - 0 19 0.0 S ksoftirqd/3
36 TS - 0 19 0.0 S ksoftirqd/4
42 TS - 0 19 0.0 S ksoftirqd/5
48 TS - 0 19 0.0 S ksoftirqd/6
54 TS - 0 19 0.0 S ksoftirqd/7
grep RTIRQ_NAME_LIST /etc/default/rtirq
# RTIRQ_NAME_LIST="rtc snd usb i8042" # old
RTIRQ_NAME_LIST="snd usb i8042" Before you boot the lowlatency kernel or a
generic kernel with
boot option 'threadirqs', make USB the head and remove the keyboard
from this list, IOW change it to RTIRQ_NAME_LIST="usb snd"
could you
elaborate how? 
20 Jul
20 Jul
10:25 a.m.
New subject: reducing xruns (System configuration)
On 07/19/2018 05:00 AM, Ralf Mardorf wrote:
>> What's the output of
>>
>> uname -rm
>
> 4.17.8-041708-lowlatency x86_64
Back to the basic issue original poster has. Here's my situation and
something I think is related.
On my desktop, I have Debian Testing. (Debian can't be described as a
'third-party' kernel.) I have their 4.16.02-2-rt-amd64 kernel running there.
Regardless of which audio device I use (USB or built-in) simply running
JACK generates at least an xrun per second. Constantly. On the previous
kernel (4.8 something from Ubuntu Studio), I ran with latency of about
1-3 msecs without xruns.
Even boosting latency up to 64msec latency doesn't change the constant
xruns.
The box has four spinning metal hard disks in it. Disk I/O has slowed
down noticeably. Network access has slowed down a lot. Even displaying
images on screen has slowed down visibly!
On my laptop, with Debian's 4.8.0-1-amd64 (non-RT kernel), I can run all
day with the same USB card at 10.7msec latency, without a single xrun.
No network or video slowdown.
4.8 doesn't have the Spectre/Meltdown patches. 4.16 does. So I blame the
Spectre/Meltdown patches.
--
David W. Jones
gnome(a)hawaii.rr.com
authenticity, honesty, community
http://dancingtreefrog.com
10:43 a.m.
New subject: reducing xruns (System configuration)
david <gnome(a)hawaii.rr.com> writes:
On my laptop, with Debian's 4.8.0-1-amd64 (non-RT
kernel), I can run
all day with the same USB card at 10.7msec latency, without a single
xrun. No network or video slowdown.
4.8 doesn't have the Spectre/Meltdown patches. 4.16 does. So I blame
the Spectre/Meltdown patches.
IIRC, the most likely culprit (page table isolation, making for quite
more expensive context switches) can be disabled on the kernel command
line at boot. This would be
pti=off
Of course, not to be used on a system providing non-privileged accounts
to third parties. But for an audio workstation, this might be
reasonable.
--
David Kastrup
12:17 p.m.
New subject: reducing xruns (System configuration)
On Thu, 19 Jul 2018 20:25:42 -1000, david wrote:
4.8 doesn't have the Spectre/Meltdown patches. 4.16
does. So I blame
the Spectre/Meltdown patches.
I already pointed tis out by an earlier mail:
"I didn't test, if the page table isolation stuff has got impact on
real-time performance, maybe you want to disable it, if so add 'nopti':
[rocketmouse at archlinux ~]$ grep -A3
'\^nopti' /boot/syslinux/syslinux.cfg MENU LABEL Arch Linux threadirqs
^nopti LINUX ../vmlinuz-linux
APPEND root=LABEL=archlinux ro threadirqs nopti
INITRD ../intel-ucode.img,../initramfs-linux.img" -
https://lists.linuxaudio.org/archives/linux-audio-user/2018-July/110576.html
By the kernel package's changelog you could see if 'nopti' is
available. For older official Ubuntu kernels is was 'nokaiser', but for
all current official Ubuntu kernels it is 'nopti'.
The output of
ls -hAl /sys/devices/system/cpu/vulnerabilities/; cat
/sys/devices/system/cpu/vulnerabilities/*
shows if PTI mitigation is enabled at all.
12:52 p.m.
New subject: reducing xruns (System configuration)
On 07/19/2018 10:17 PM, Ralf Mardorf wrote:
On Thu, 19 Jul 2018 20:25:42 -1000, david wrote:
And I figured out where to put pti=off on my various GRUB lines, so now
it boots with it off. And my desktop latency is back to 10msecs. :)
My understanding is that any post-4.13.x kernel has the Spectre/Meltdown
patches. I first tried it with 4.13.x and decided that items like random
kernel panios, system freezes and crashes weren't very good ways to
defend against Spectre/Meltdown/DOS. ;)
--
David W. Jones
gnome(a)hawaii.rr.com
authenticity, honesty, community
http://dancingtreefrog.com
4.8 doesn't have the Spectre/Meltdown
patches. 4.16 does. So I blame
the Spectre/Meltdown patches.
I already pointed tis out by an earlier mail:
"I didn't test, if the page table isolation stuff has got impact on
real-time performance, maybe you want to disable it, if so add 'nopti':
[rocketmouse at archlinux ~]$ grep -A3
'\^nopti' /boot/syslinux/syslinux.cfg MENU LABEL Arch Linux threadirqs
^nopti LINUX ../vmlinuz-linux
APPEND root=LABEL=archlinux ro threadirqs nopti
INITRD ../intel-ucode.img,../initramfs-linux.img" -
https://lists.linuxaudio.org/archives/linux-audio-user/2018-July/110576.html
By the kernel package's changelog you could see if 'nopti' is
available. For older official Ubuntu kernels is was 'nokaiser', but for
all current official Ubuntu kernels it is 'nopti'.
The output of
ls -hAl /sys/devices/system/cpu/vulnerabilities/; cat
/sys/devices/system/cpu/vulnerabilities/*
shows if PTI mitigation is enabled at all.
6:21 p.m.
New subject: reducing xruns (System configuration)
On Fri, July 20, 2018 3:52 am, david wrote:
And I figured out where to put pti=off on my various
GRUB lines, so now
it boots with it off. And my desktop latency is back to 10msecs. :)
That is good to know. And depressing, one more thing to add to the latency
performance checklist. Set RT permissions, add user to audio group, etc.
...oh, and one last thing, you need to decide whether you want better
performance or want to protect against malicious software reading you
kernel memory. You do have separate dedicated computers for audio and
banking, right? And a third for general web use? Just make sure you
banking and web browsing computers leave PTI enabled, it's probably OK to
disable it on your dedicated audio computer since you don't connect it to
the Internet.
--
Chris Caudle
6:31 p.m.
New subject: reducing xruns (System configuration)
"Chris Caudle" <chris(a)chriscaudle.org> writes:
On Fri, July 20, 2018 3:52 am, david wrote:
AMD could probably make a killing by putting out old processors with
Sandy/Ivy bridge compatible behavior. Pity that likely the most
incurable incompatibility would be the BIOS for the integrated GPU
(which is already blocking upgrades of Thinkpad T420 and their ilk to
Ivy bridge).
--
David Kastrup
And I figured out where to put pti=off on my
various GRUB lines, so now
it boots with it off. And my desktop latency is back to 10msecs. :)
That is good to know. And depressing, one more thing to add to the
latency performance checklist. Set RT permissions, add user to audio
group, etc. ...oh, and one last thing, you need to decide whether you
want better performance or want to protect against malicious software
reading you kernel memory. You do have separate dedicated computers
for audio and banking, right? And a third for general web use? Just
make sure you banking and web browsing computers leave PTI enabled,
it's probably OK to disable it on your dedicated audio computer since
you don't connect it to the Internet.
6:59 p.m.
New subject: reducing xruns (System configuration)
On Fri, 20 Jul 2018 09:21:07 -0500, Chris Caudle wrote:
You do have separate dedicated computers for audio and
banking, right?
Not everybody does online banking during a real-time audio session. If
the computer is disconnected from the Internet during an audio
session, with Meltdown mitigation disabled, nobody could access the
computer via the Internet. If the real-time audio session is finished,
just reboot the computer with page table isolation enabled, connect to
the Internet and do your online-banking.
Some users might be connected to the Internet with an insecure
computer, because security isn't something they are concerned about.
IMO security is important, but OTOH I never ever would wear a wallet
connected to a chain.
See https://www.wikihow.com/Secure-a-Wallet .
1. This isn't secure at all.
2. Even if it should be secure, you would pay the price to look like an
idiot.
Safety measures should be adequate, for your wallet, as well as for
your computer. If a safety measure should be inadequate, .e.g. because
you look like an idiot or because it has got negative impact on your
computer usage, there might be other safety measures to reach identical
good security. How about router with a firewall and other safety
measures (I'm not an expert)? The router should use page table
isolation, the computers connected to the router perhaps don't need it
anymore.
7:48 p.m.
New subject: reducing xruns (System configuration)
Ralf Mardorf <ralf.mardorf(a)alice-dsl.net> writes:
Some users might be connected to the Internet with an
insecure
computer, because security isn't something they are concerned about.
IMO security is important, but OTOH I never ever would wear a wallet
connected to a chain.
See https://www.wikihow.com/Secure-a-Wallet .
1. This isn't secure at all.
2. Even if it should be secure, you would pay the price to look like an
idiot.
Shrug. I use a watch chain after having lost pocket watches to the
floor. The flywheel axle just is too brittle and you cannot afford
replacements. I'd rather look like an idiot than be one.
Safety measures should be adequate, for your wallet,
as well as for
your computer. If a safety measure should be inadequate, .e.g. because
you look like an idiot or because it has got negative impact on your
computer usage, there might be other safety measures to reach
identical good security. How about router with a firewall and other
safety measures (I'm not an expert)? The router should use page table
isolation, the computers connected to the router perhaps don't need it
anymore.
You are aware that JavaScript exploits have already been created? A
Firewall will not protect you from looking at websites. Or if it does,
it's likely easier to just not connect in the first place rather than
use a Firewall that is essentially a brick.
--
David Kastrup
8:59 p.m.
New subject: reducing xruns (System configuration)
On Fri, 20 Jul 2018 17:48:45 +0200, David Kastrup wrote:
I use a watch chain after having lost pocket watches to
the
floor. The flywheel axle just is too brittle and you cannot afford
replacements. I'd rather look like an idiot than be one.
A watch chain is something completely different, since it's
unostentatiously filigree and it's worn by completely different people,
for exactly the purpose, to prevent against dropping the clock. Sure,
sometimes it's jewellery, an earring is jewellery, too, there's nothing
wrong with this.
A wallet chain cannot be overlooked and is warn by not to different
kind of men, in combination with additional would-be masculinely
clothes and absurd thick wallets. Their earrings are more likely tunnels
and their is nothing wrong with a tunnel, unless the tunnel isn't
overdone, but here we go, the purpose is to be just somebody of a
group, but within this group, the most extrem accepted of the group.
However, in the context of the analogy, it is a picture for
"proportionality". I don't know if there are other safety measures that
could make page table isolation unneeded, other than the one I
described. If page table isolation should have bad impact on real-time
audio sessions, than disconnect from the Internet and reboot with
'nopti'. For online usage reboot with page table isolation enabled,
before you connect to the Internet. No second machine is required, it
could be done by one computer with at least two boot menu entries.
Maybe, I don't know, there are other workarounds for those who want to
use the Internet during an audio session, with the same computer. If it
shouldn't be a router with a firewall, perhaps something else.
In a nutshell, if the OP wants to use an USB audio Interface with this
particular IC, at that low latency, I suspect taking care about
Meltdown JavaScript exploits isn't an option.
From a mail the OP wrote:
> PID CLS RTPRIO NI PRI %CPU STAT COMMAND
> 441 FF 50 - 90 0.0 S irq/29-mei_me
> 527 FF 50 - 90 0.0 S irq/31-iwlwifi
> 8 TS - 0 19 0.0 S ksoftirqd/0
> 18 TS - 0 19 0.0 S ksoftirqd/1
> 24 TS - 0 19 0.0 S ksoftirqd/2
> 30 TS - 0 19 0.0 S ksoftirqd/3
> 36 TS - 0 19 0.0 S ksoftirqd/4
> 42 TS - 0 19 0.0 S ksoftirqd/5
> 48 TS - 0 19 0.0 S ksoftirqd/6
> 54 TS - 0 19 0.0 S ksoftirqd/7
The OP anyway should get rid of the wlan adapter. Maybe PTI isn't an
issue at all, but if it should be, it shouldn't matter.
21 Jul
21 Jul
10:52 a.m.
New subject: reducing xruns (System configuration)
On 07/20/2018 04:21 AM, Chris Caudle wrote:
On Fri, July 20, 2018 3:52 am, david wrote:
Mozilla implemented Meltdown/Spectre patches beginning in v57.0.4:
https://news.softpedia.com/news/mozilla-firefox-57-0-4-released-with-meltdo…
And I always run with Javascript disabled plus adblocking to keep out
unwanted intrusions on my browsing experience. ;)
--
David W. Jones
gnome(a)hawaii.rr.com
authenticity, honesty, community
http://dancingtreefrog.com
And I figured out where to put pti=off on my
various GRUB lines, so now
it boots with it off. And my desktop latency is back to 10msecs. :)
That is good to know. And depressing, one more thing to add to the latency
performance checklist. Set RT permissions, add user to audio group, etc.
...oh, and one last thing, you need to decide whether you want better
performance or want to protect against malicious software reading you
kernel memory. You do have separate dedicated computers for audio and
banking, right? And a third for general web use? Just make sure you
banking and web browsing computers leave PTI enabled, it's probably OK to
disable it on your dedicated audio computer since you don't connect it to
the Internet.
2262
days inactive
2265
days old
linux-audio-user@lists.linuxaudio.org
14 comments
5 participants
participants (5)
-
Chris Caudle -
david -
David Kastrup -
Max -
Ralf Mardorf