Horrific jackd xrun behaviour after upgrade from US 10.04 to 12.04 (Re: R: Audio issues in 10.04)
Am Mon, 17 Sep 2012 15:30:49 +0200 schrieb Thomas Orgis thomas-fo...@orgis.org: Please post the output of $ /etc/init.d/rtirq status Will try to get to that tonight. Well ... this is interesting. I tried AVLinux and this also showed massive xruns, with the added twist of ffado giving up on the device after some time. I then thought that maybe swapping the firewire card to another slot might help. It shared IRQ with some USB ports. I shoved it into the x16 slot (using onboard graphics). Now, would I have been properly awake at that time, I'd have thought that the generally abysmal performance cannot have something to do with the firewire card, as it also shows for internal audio and a USB card. But, perhaps it was a step towards the final goal. Thing is, the machine did not want to start after swapping the firewire card. Pulling the card out, waiting, connecting/disconnecting power ... and the machine boots again. To shorten it a bit: I must have had something funky with the power connection to the card. On the same wire branch s a SATA power adapter for the 3rd hard disk and I suspect it having had some bad insulation, causing flaky current / shorting out. That would explain the box dying on moving said branch about a bit, and some protection circuit preventing powering on again (for some time). I managed to get the thing running again after adding insulation to the supposedly faulty adapter, and what might play a part: The BIOS got reset. I'm with defaults there now. I'm leaving the box running for the day with JACK on firewire under UbuntuStudio (my partial KXStudio). During the first minutes I did not see one underrun and could record a bit with Ardour. II faintly remember that I checked internal audio, too, and it seemed to be fine. So it could be down to flaky power to the firewire card / the system at all (but how could that subtly influence the machine without killing it outright?) or, what I think is more probable, some BIOS setting. I'll check that later, if the test is successful. Anoter remote possibility would be that the machine freaks out generally when the first PCIe slot is occupied --- I'll move the firewire card back to see if that makes a change. Anyhow, here's the current interrupts (see how the one for firewire is the most populated?): CPU0 CPU1 CPU2 0:124 0 1 IO-APIC-edge timer 1:465 0 25 IO-APIC-edge i8042 7: 1 0 0 IO-APIC-edge parport0 8: 0 0 1 IO-APIC-edge rtc0 9: 0 0 0 IO-APIC-fasteoi acpi 12: 12464 19 1858 IO-APIC-edge i8042 14: 0 0 0 IO-APIC-edge pata_atiixp 15: 0 0 0 IO-APIC-edge pata_atiixp 16: 0 2838 IO-APIC-fasteoi ohci_hcd:usb3, ohci_hcd:usb4, snd_hda_intel 17: 1 1682 2187 IO-APIC-fasteoi ehci_hcd:usb1 18: 50797 4975 IO-APIC-fasteoi ohci_hcd:usb5, ohci_hcd:usb6, ohci_hcd:usb7, radeon, firewire_ohci 19: 0 5 1747 IO-APIC-fasteoi ehci_hcd:usb2 22: 1442 60 15896 IO-APIC-fasteoi ahci 42: 0 0 0 PCI-MSI-edge eth3 NMI: 0 0 0 Non-maskable interrupts LOC: 215017 224583 218795 Local timer interrupts SPU: 0 0 0 Spurious interrupts PMI: 0 0 0 Performance monitoring interrupts IWI: 0 0 0 IRQ work interrupts RES: 38575 50657 61704 Rescheduling interrupts CAL:498451613 Function call interrupts TLB:759889843 TLB shootdowns TRM: 0 0 0 Thermal event interrupts THR: 0 0 0 Threshold APIC interrupts MCE: 0 0 0 Machine check exceptions MCP: 1 1 1 Machine check polls ERR: 1 MIS: 0 rtirq: PID CLS RTPRIO NI PRI %CPU STAT COMMAND 302 FF 90 - 130 0.6 Sirq/18-firewire 994 FF 83 - 123 0.0 Sirq/16-snd_hda_ 84 FF 80 - 120 0.0 Sirq/17-ehci_hcd 88 FF 80 - 120 0.0 Sirq/16-ohci_hcd 86 FF 79 - 119 0.0 Sirq/19-ehci_hcd 90 FF 79 - 119 0.0 Sirq/16-ohci_hcd 99 FF 75 - 115 0.0 Sirq/1-i8042 98 FF 74 - 114 0.1 Sirq/12-i8042 23 FF 50 - 90 0.0 Sirq/9-acpi 73 FF 50 - 90 0.4 Sirq/22-ahci 92 FF 50 - 90 0.1 Sirq/18-ohci_hcd 94 FF 50 - 90 0.1 Sirq/18-ohci_hcd 96 FF 50 - 90 0.1 Sirq/18-ohci_hcd 100 FF 50 - 90 0.0 Sirq/8-rtc0 287
Re: Horrific jackd xrun behaviour after upgrade from US 10.04 to 12.04 (Re: R: Audio issues in 10.04)
On Tue, 2012-09-18 at 11:39 +0200, Thomas Orgis wrote: rtirq: PID CLS RTPRIO NI PRI %CPU STAT COMMAND 302 FF 90 - 130 0.6 Sirq/18-firewire 994 FF 83 - 123 0.0 Sirq/16-snd_hda_ 84 FF 80 - 120 0.0 Sirq/17-ehci_hcd 88 FF 80 - 120 0.0 Sirq/16-ohci_hcd 86 FF 79 - 119 0.0 Sirq/19-ehci_hcd 90 FF 79 - 119 0.0 Sirq/16-ohci_hcd 99 FF 75 - 115 0.0 Sirq/1-i8042 98 FF 74 - 114 0.1 Sirq/12-i8042 23 FF 50 - 90 0.0 Sirq/9-acpi 73 FF 50 - 90 0.4 Sirq/22-ahci 92 FF 50 - 90 0.1 Sirq/18-ohci_hcd 94 FF 50 - 90 0.1 Sirq/18-ohci_hcd 96 FF 50 - 90 0.1 Sirq/18-ohci_hcd 100 FF 50 - 90 0.0 Sirq/8-rtc0 287 FF 50 - 90 0.0 Sirq/14-pata_ati 288 FF 50 - 90 0.0 Sirq/15-pata_ati 295 FF 50 - 90 0.1 Sirq/18-radeon 965 FF 50 - 90 0.0 Sirq/7-parport0 1212 FF 50 - 90 0.0 Sirq/42-eth3 3 TS - 0 19 1.0 Rksoftirqd/0 9 TS - 0 19 1.0 Sksoftirqd/1 13 TS - 0 19 1.0 Sksoftirqd/2 I'm short in time and can't read your email right now, so I only take a look at the output of rtirq status. Usually you can't do anything against the shared IRQ, however it's ok, since firewire and snd_hda are head of the IRQs. But the first IRQ, before firewire has to be the timer rtc0. Change the config to RTIRQ_NAME_LIST=rtc [and add anything as it was before, excepted of a second rtc entry] Hth, Ralf -- Ubuntu-Studio-users mailing list Ubuntu-Studio-users@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-studio-users
Re: Horrific jackd xrun behaviour after upgrade from US 10.04 to 12.04 (Re: R: Audio issues in 10.04)
I'm short in time and can't read your email right now, so I only take a look at the output of rtirq status. Usually you can't do anything against the shared IRQ, however it's ok, since firewire and snd_hda are head of the IRQs. But the first IRQ, before firewire has to be the timer rtc0. Change the config to RTIRQ_NAME_LIST=rtc [and add anything as it was before, excepted of a second rtc entry] PS: After editing the config stop and start might not do the job and a reboot could be required, restart usually won't work too, perhaps reset and force-reload will work. However, a reboot is safe. -- Ubuntu-Studio-users mailing list Ubuntu-Studio-users@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-studio-users
Re: Horrific jackd xrun behaviour after upgrade from US 10.04 to 12.04 (Re: R: Audio issues in 10.04)
Am Tue, 18 Sep 2012 12:37:21 +0200 schrieb Ralf Mardorf ralf.mard...@alice-dsl.net: Change the config to RTIRQ_NAME_LIST=rtc [and add anything as it was before, excepted of a second rtc entry] PS: After editing the config stop and start might not do the job and a reboot could be required No biggie ... I'll bear with some reboots when I get the thing stable afterwards. As for rtc being first in the list: If that is necessary (I could imagine that it is), I wonder why that is not default on UbuntuStudio (I remember the update modifying that config anyway). And, having read up a bit ... AHCI seems to be known to cause audio issues, also on Windows machines (where people endure greate pain to switch between AHCI and IDE mode ... registry hacking, or even reinstall ... I didn't have to do a thing with Linux and I don't see why I should). Is this a consensus here: If building a DAW, stay away from AHCI? Though, I think that I also had it enabled in 10.04 (on another mainboard, though); that could be a regression in the kernel behaviour, which would be depressing (I'm getting old, as I see lots of regressing with Linux in general, especially considering reliability). Alrighty then, Thomas signature.asc Description: PGP signature -- Ubuntu-Studio-users mailing list Ubuntu-Studio-users@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-studio-users
Horrific jackd xrun behaviour after upgrade from US 10.04 to 12.04
Hi, the subject says it all ... I had an install of 10.04 that worked. Only issue seemed to be instability when recording while a USB drive is connected. So, considering that perhaps this is a sign to replace a mainboard with broken onboard USB (used a PCI add-on card, which might have worsened the interrupt situation), I did replace it and the machine worked fine during a session. Now with the fully replaced hardware (Asus AM3 board with 780G chipset, Athlon || X3 460, 8 Gig of Kingston DDR3 ValueRAM, 250 G Seagate SATA boot, 2 WD EADS on mdraid for recordings and not to forget VIA firewire on PCI-E (same card as before)) and the update to ubuntu 12.04 (yes, should have tested the final hardware with 10.04 first, eh?), the performance is hindered by jackd not being able to keep steady without generating xruns at a some rate. Not really constant rate, though, also its behaviour depends on client connections (even when just connecting meterbridge, this seems to help triggering xruns a lot). A very interesting fact is that using 3x512 periods (or bigger) is less stable even than going down to 3x32! With big periods, I get xruns right away, while with the low setting, I was able to get an hour of recording done, but that ended prematurely -- I _guess_ that this was because of some software glitch (like xrun handling) and not due to the bass player nudging the keyboard by chance. But I cannot be sure about that. Now, I do have the lowlatency kernel already installed, also fresh jack/ardour from kxstudio ... have rtirq setup updated by dpkg (firewire in there instead of ohci1394). What are the ubuntu studio folks' thoughts on this? Did you encounter _more_ stable jack with extremely low latencies? But since it is not really stable and glitch-free in any config, this interesting characteristic does not help. Oh, and it happens independent of cpufreq governor. I do use XFCE and the integrated radeon with open source driver. Any help on getting that setup stable again is appreciated ... or should I simply go back to 10.04 (and hand-install current ardour/jack, as I did before)? I figure that I shouldn't even need a lowlat kernel for getting basic 3x512 recording work! Alrighty then, Thomas. PS: Why upgrade at all? Well, I have always a spark of hope that some iteration of the GNU/Linux audio ecosystem will be really stable, without random crashes of Ardour, for example. But I guess one has to live with crashing multimedia apps ... not been that different during my days doing video with Ulead Media Studio on Windows (and the fact that version 5 was less usable than 2.5). signature.asc Description: PGP signature -- Ubuntu-Studio-users mailing list Ubuntu-Studio-users@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-studio-users
Re: Horrific jackd xrun behaviour after upgrade from US 10.04 to 12.04
PS: CPU frequency scaling? -- Ubuntu-Studio-users mailing list Ubuntu-Studio-users@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-studio-users
Re: Horrific jackd xrun behaviour after upgrade from US 10.04 to 12.04
Am Mon, 17 Sep 2012 15:15:58 +0200 schrieb Ralf Mardorf ralf.mard...@alice-dsl.net: PS: CPU frequency scaling? I noted that it is independent of that. I switched to 'performance' governor and nothing changed. Also, my experience with working setups is that frequency scaling does not matter to them. That might be exceptional, but I remember reliable firewire recording with my old Thinkpad X31, using freq scaling and tickless kernel (back then when that was a new feature;-) While I will do tests with scaling disabled, I sincerely hope that having it on will not interfere. Heck, even on 'powersave' it'll still be frikkin' 3 cores at around/over 1 GHz or so;-) Alrighty then, Thomas signature.asc Description: PGP signature -- Ubuntu-Studio-users mailing list Ubuntu-Studio-users@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-studio-users
