Bug 65231

Summary:	[BISECTED]workqueue leaked lock or atomic/scheduling while atomic in hid_magicmouse or bluetooth
Product:	Drivers	Reporter:	Dmytro (edio)
Component:	Bluetooth	Assignee:	linux-bluetooth (linux-bluetooth)
Status:	NEW ---
Severity:	high	CC:	alan, chris.kennedy, dh.herrmann, dmitry.torokhov, linux-kernel
Priority:	P1
Hardware:	x86-64
OS:	Linux
Kernel Version:	3.12.0-1-ARCH	Subsystem:
Regression:	Yes	Bisected commit-id:
Attachments:	dmesg output after Trackpad is connected

Description Dmytro 2013-11-19 22:58:18 UTC

Connecting Apple Magic Trackpad causes system freeze after few seconds of usage.

Hardware:
Broadcom based bluetooth adapter: bcm2045

Related software:
bluez 4.101

Comment 1 Dmytro 2013-11-19 23:01:57 UTC

Created attachment 115161 [details]
dmesg output after Trackpad is connected

Comment 2 Dmytro 2013-11-23 20:31:52 UTC

Not reproduced in kernel 3.10.20 (lts)

Comment 3 Alan 2013-11-26 22:00:25 UTC

[ 2723.505055] BUG: workqueue leaked lock or atomic: kworker/u9:0/0xfffffffe/1692
     last function: hci_rx_work [bluetooth]
[ 2723.505062] CPU: 1 PID: 1692 Comm: kworker/u9:0 Tainted: P           O 3.12.0-1-ARCH #1
[ 2723.505065] Hardware name: System manufacturer P5K/P5K, BIOS 1006    03/05/2008


Please verify you can duplicate this bug on a system where you've not loaded the proprietary Nvidia module. I suspect it'll show up the same but please confirm that

Comment 4 Dmytro 2013-11-27 22:08:08 UTC

(In reply to Alan from comment #3)

Hi, Alan.

Just checked on the same system with nouveau instead of nvidia.

Behaviour is slightly different:
I don't get those error messages when moving cursor, but I do get system freezes each time I swipe touchpad with 2 fingers.

Unfortunately, effect is so instant, that I'm not able to capture dmesg output. Nor I able to find any traces of those freezes in journactl after reboot.

Comment 5 Mathias 2014-01-06 09:47:23 UTC

Just wanted to add that I have the same issue (bluez 5, no nvidia hardware) and that I've seen it reported else-where:

https://bbs.archlinux.org/viewtopic.php?id=173369
https://bugs.archlinux.org/task/38290
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1248287

the following commit seems to cause the problem. If I revert it it works again without kernel panic:
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/?id=b1a1442a23776756b254b69786848a94d92445ba

Comment 6 Alan 2014-01-06 11:58:57 UTC

This actually looks like bluetooth and magicmouse are the problem

Comment 7 David Herrmann 2014-01-06 12:21:52 UTC

The bug is known and there is a fix pending. It's a buffer-overflow in HID core. Waiting for Jiri's comments:
http://www.spinics.net/lists/linux-bluetooth/msg41725.html