apple's mighty mouse gets a lot of USB disconnects during MTP probing

Thank you for your bug report, it seems another case where the mtp option takes over devices on rhythmbox

Could you get a "udevadm info --exportdb" log there?

nope, I don't own that mouse

Just for referenct, if somebody else wants to provide such log, the right command is
udevadm info --export-db

I just experienced the same problem after upgrading to Lucid, but with a Kensington mouse. The mouse stopped working until I killed rhythmbox.

[ 213.820038] usb 4-2: new low speed USB device using uhci_hcd and address 26
[ 214.002066] usb 4-2: configuration #1 chosen from 1 choice
[ 214.018342] input: Kensington USB/PS2 Wheel Mouse as /devices/pci0000:00/0000:00:1a.1/usb4/4-2/4-2:1.0/input/input31
[ 214.018438] generic-usb 0003:047D:1029.001B: input,hidraw2: USB HID v1.10 Mouse [Kensington USB/PS2 Wheel Mouse] on usb-0000:00:1a.1-2/input0
[ 214.042558] usb 4-2: USB disconnect, address 26
[ 214.122753] usb 4-2: usbfs: USBDEVFS_CONTROL failed cmd rhythmbox rqt 128 rq 6 len 1024 ret -108

Disabling the MTP plugin in Rhythmbox fixes the problem. I don't have any MTP devices, so this is an acceptable work-around for me. I used gconf-editor to disable the plugin.

With that mouse being plugged in, can you please copy&paste the output of "mtp-detect" (it's in the "mtp-tools" package, you have to install it).

Also, can you please start rhythmbox in debug mode, so that we can see what it's doing? For this, please run

  rhythmbox -D mtp > /tmp/rb.log 2>&1

Wait until Rhythmbox displays and settles down, then close it again and attach /tmp/rb.log here. Thank you!

> mtp-detect

libmtp version: 1.0.2

Listing raw device(s)
   No raw devices found.

The USB ID for the mouse is 4-2.

So, RB does not actually try to talk to it as a music player, but it seems the act of checking _whether_ it is an MTP player already distracts the mouse.

Can you please do the following: Close RB, detach the mouse, run

  udevadm monitor --udev -e

then attach the mouse and let the output settle. Copy&paste all output (you probably have to scroll up). Now detach it again, open Rhythmbox, press Enter some 10 times (so that you get enough free space in the Terminal to see where the new output starts), attach the mouse again. Let it settle, and copy&paste all the new output.

Finally, please attach /var/log/kern.log and /var/log/Xorg.0.log.

Thanks!

Ah, thanks for the logs. The kernel log shows that the mouse bounces a lot during this, there are tons of USB reconnects.

So it seems libmtp does something funky with the device that causes it to bounce.

Can you please start rhythmbox without the mouse attached, then open a Terminal and do

   strace -vvfs1024 -o /tmp/strace.txt -p `pidof rhythmbox`

then plug in the mouse, wait a little, then press Control-C. Please attach /tmp/strace.txt here. Thanks!

Same problem with a "RollerMouse Pro".
Attaching a no-name usb mouse as a second mouse works and I can use the application. The RollerMouse Pro works again after Rhythmbox is closed down.

Happens also with a BENQ optical mouse (M101 model).
Thanks a lot for the workaround. If you need any more details, do not hesitate!

Thanks a lot for the strace. This revealed something interesting which might point out the root cause:

15840 open("/dev/bus/usb/006/002", O_RDWR) = 42
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3b9e0) = 4
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3b9e0) = 4
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3bb00) = -1 EILSEQ (Invalid or incomplete multibyte or wide character)

Assuming that you attached the mouse to the same USB port as at the time when you supplied the udev and kernel logs, I believe that this is the mouse. Probing the next device then apparently causes the USB reset:

15840 close(42) = 0
15840 open("/dev/bus/usb/004/002", O_RDWR) = 42
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3b9e0) = 4
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3b9e0) = 4
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3b9e0) = 4
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3b9e0) = 4
15840 ioctl(42, USBDEVFS_CONTROL, 0x7fffd8f3bb00) = -1 EPIPE (Broken pipe)

at which point the kernel re-detects the devices, RB re-probes all USB devices, and the cycle starts again.

I'll forward this to upstream with a few thoughts.

This bug was fixed in the package libmtp - 1.0.3-1ubuntu1

---------------
libmtp (1.0.3-1ubuntu1) maverick; urgency=low

  * Merge with debian unstable (LP: #585802), Ubuntu remaining changes:
    - Install the apport package hook.

libmtp (1.0.3-1) unstable; urgency=low

  * debian/copyright: Update according to DEP-5 proposal.
  * Add apport package hook to reduce delta with Ubuntu.
  * New upstream release:
    - Don't probe HID equipment for device "OS descriptors"
      some Kensington mice really don't like that (LP: #559892).
  * Switch to debian 3.0 (quilt) source format.
 -- Alessio Treglia <email address hidden> Wed, 26 May 2010 14:01:48 +0200

Apart from freezing my Apple Mighty Mouse now the Rhythmox window freezes itself (turns gray) and stays that way until I force close it. If it's any help I attach here output of the command listed above:

rhythmbox -D mtp > /tmp/rb.log 2>&1

NB this happens when Banshee player is running, i.e. playing music.