[LAU] Disabling IRQ #19

[Peter P.]

Dear List,

I have a strange and unreproducable error on my laptop, that halts the
entire audio system randomly about once every three weeks.

Jackd quits, with it all clients, and dmesg says:

[21684.947293] irq 19: nobody cared (try booting with the "irqpoll"option) 
[21684.947303] Pid: 194, comm: irq/19-ehci_hcd Not tainted 3.2.0-4-rt-amd64 #1 Debian 3.2.54-2
[21684.947309] Call Trace:
[21684.947327]  [<ffffffff81098889>] ? __report_bad_irq+0x2c/0xb5
[21684.947336]  [<ffffffff81098c6d>] ? note_interrupt+0x16f/0x1f2
[21684.947344]  [<ffffffff81097969>] ? irq_thread_fn+0x32/0x32
[21684.947351]  [<ffffffff81097969>] ? irq_thread_fn+0x32/0x32
[21684.947358]  [<ffffffff8109783c>] ? irq_thread+0x106/0x201
[21684.947367]  [<ffffffff81097736>] ? irq_finalize_oneshot+0xb3/0xb3
[21684.947378]  [<ffffffff81062ee4>] ? kthread+0x78/0x80
[21684.947385]  [<ffffffff8103f2fa>] ? get_parent_ip+0x9/0x1b
[21684.947395]  [<ffffffff8136a0f4>] ? kernel_thread_helper+0x4/0x10
[21684.947405]  [<ffffffff81062e6c>] ?  rcu_read_unlock_sched_notrace+0x2a/0x2a
[21684.947412]  [<ffffffff8136a0f0>] ? gs_change+0x13/0x13
[21684.947417] handlers:
[21684.947429] [<ffffffff81096e34>] irq_default_primary_handler threaded [<ffffffffa003d483>] usb_hcd_irq
[21684.947473] [<ffffffff81096e34>] irq_default_primary_handler threaded [<ffffffffa025a85a>] ips_irq_handler
[21684.947485] [<ffffffff81096e34>] irq_default_primary_handler threaded [<ffffffffa04d62cd>] snd_hdsp_interrupt
[21684.947498] Disabling IRQ #19

The sound card is an RME Multiface via an ExpressCard in a Thinkpad
X201s. The card itself had been working reliably since more than ten
years by now (not with the ExpressCard though). Running on Debian RT
Kernel.

I have no way to set/change the IRQ assignments in this BIOS.
Interrupt 19 is shared by modules ehci_hcd:usb2, ips, snd_hdsp,
where the two others are USB and the thermal subsystem.
I assume ehci_hcd:usb2 denotes the second USB bus. If this is correct,
there are no devices connected to it at the moment, according to lsusb
(except for USB hubs).

Web research has shown that people suggest using the kernel option
"irqpoll" or disable "threadirqs", use "noirqdebug" or "irqdebug" options
(apparently quite expensive in computation), or the irqbalance
package.

https://bbs.archlinux.org/viewtopic.php?id=133327 gives a hint about
deinstalling laptop-mode-tools, which I am running, and which made the
problem go away for the poster.

As the error occurs so infrequently, I can not yet verify if any of
these help, and am hoping for some hints or experience from this list.

Please, any help or ideas are desperately needed, as this is a
production system that is used in live shows in front of large
audiences, and I absolutely need to resolve that error.

Thank you for all ideas,
best, Peter