[PlanetCCRMA] USB audio interface kills uhci_hcd module

[Luis Garrido]

Hi!

I've been running the 2.6.26.8-1.rt16.1.fc10.ccrma.i686 rt kernel for
a while without any complaints. Only recently I had the need to use my
USB audio interface Edirol UA-25 for a recording and found a strange
bug.

A few seconds after pluging it in the uhci subsystem drops dead,
taking the whole USB stack down with it:

Aug 20 13:50:35 membrillo kernel: uhci_hcd 0000:00:1d.3: host
controller process error, something bad happened!
Aug 20 13:50:35 membrillo kernel: uhci_hcd 0000:00:1d.3: host
controller halted, very bad!
Aug 20 13:50:35 membrillo kernel: uhci_hcd 0000:00:1d.3: HC died; cleaning up
Aug 20 13:50:35 membrillo kernel: usb 5-1: USB disconnect, address 2

For all the vaunted Linux rock solid stability, if I happen to have
jackd running attached to it the effect will be the same than if I
unplug the USB module: the system will freeze beyond any soft
resetting. Boo! }:-)

This won't happen with other USB devices like a pendrive or a Bluetooth dongle.

Other kernels like the Fedora stock one 2.6.27.29-170.2.79.fc10.i686
or the Planet experimental 2.6.29-0.1.rt3.1.rc6.fc10.ccrma.i686
exhibit the exact same behaviour.

In the end I was hard pressed for time and had to use another laptop
which was running a Ubuntu standard kernel. I configured a somehow
conservative latency (64 ms) and was able to record 90 minutes of
48kHz stereo with only one single xrun (and it happened when I first
launched Ardour, so it doesn't count), it is nice to confirm that
stock kernels nowadays are reliable enough from some undemanding audio
tasks.

After the session I googled around and found a possible workaround:
removing the ehci_hcd driver.

rmmod ehci_hcd

This disables high speed USB, but since the UA-25 is a USB 1.1 device
anyway I can live with that while working with audio.

I have tried it and it does seem to improve things, but after a couple
of hours the module will die anyway:

Aug 21 13:22:10 membrillo kernel: ehci_hcd 0000:00:1d.7: remove, state 4
Aug 21 13:22:10 membrillo kernel: usb usb1: USB disconnect, address 1
Aug 21 13:22:10 membrillo kernel: ehci_hcd 0000:00:1d.7: USB bus 1 deregistered
Aug 21 13:22:10 membrillo kernel: ACPI: PCI interrupt for device
0000:00:1d.7 disabled
Aug 21 13:22:23 membrillo kernel: usb 2-2: new full speed USB device
using uhci_hcd and address 2
Aug 21 13:22:23 membrillo kernel: usb 2-2: configuration #1 chosen from 1 choice
Aug 21 13:22:23 membrillo kernel: usb 2-2: New USB device found,
idVendor=0582, idProduct=0074
Aug 21 13:22:23 membrillo kernel: usb 2-2: New USB device strings:
Mfr=1, Product=2, SerialNumber=0
Aug 21 13:22:23 membrillo kernel: usb 2-2: Product: EDIROL UA-25
Aug 21 13:22:23 membrillo kernel: usb 2-2: Manufacturer: Roland
Aug 21 13:22:24 membrillo kernel: usbcore: registered new interface
driver snd-usb-audio
Aug 21 13:26:09 membrillo kernel: [drm:i915_getparam] *ERROR* Unknown
parameter 5
Aug 21 15:16:23 membrillo kernel: uhci_hcd 0000:00:1d.0: host
controller process error, something bad ha
ppened!
Aug 21 15:16:23 membrillo kernel: uhci_hcd 0000:00:1d.0: host
controller halted, very bad!
Aug 21 15:16:23 membrillo kernel: uhci_hcd 0000:00:1d.0: HC died; cleaning up
Aug 21 15:16:23 membrillo kernel: usb 2-2: USB disconnect, address 2

I don't entirely rule out possible hardware failure, I have read that
after some inevitable motherboard flexing it is not entirely uncommon
that laptop's connections with external ports suffer and may exhibit
connection problems, I had already to replace this laptop's mobo
shortly after purchasing it because of freezes and problems with the
USB ports. BUT... it seems to only happen with this particular device
and only since I upgraded from FC7 to FC10. I'll try some other live
distros and see.

Cheers,

Luis