Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

[simon]

> [  117.236007]  [<ffffffff814d2ccf>] device_del+0x18f/0x270
> [  117.236007]  [<ffffffff8109b340>] ? wake_up_q+0x70/0x70
> [  117.236007]  [<ffffffff814e97da>] _request_firmware+0x5aa/0xaf0
> [  117.236007]  [<ffffffff814e9d55>] request_firmware+0x35/0x50
> [  117.236007]  [<ffffffffc00fb881>] btbcm_setup_patchram+0x191/0x910
> [btbcm]
> [  117.236007]  [<ffffffff814e0994>] ? rpm_idle+0xc4/0x200
> [  117.236007]  [<ffffffffc0e28488>] hci_dev_do_open+0xd8/0x500

Looking between log from 3.19
--
Jul  7 21:42:57 retrobox kernel: [  107.562441] bluetooth hci0: Direct
firmware load
for brcm/BCM20702A0-0a5c-21e8.hcd failed with error -2
Jul  7 21:42:57 retrobox kernel: [  107.562452] Bluetooth: hci0: BCM: patch
brcm/BCM20702A0-0a5c-21e8.hcd not found
--

And the log of the lockup:
https://www.flickr.com/photos/24244464@N03/19375918529/sizes/o/

It looks like the firmware 'opt_flags' must be different, so this may be a
contributing factor.

In fact I found a log from 4.1.0rc7, which shows they recently changed!!
--
Jul 15 21:17:40 blind-fury kernel: [    0.000000] Linux version 4.1.0-rc7+
(root@blind-fury) (gcc version 4.9.2 (Ubuntu 4.9.2-10ubuntu13) ) #2 SMP
Wed Jun 10 21:25:17 MDT 2015
Jul 15 21:17:40 blind-fury kernel: [    0.000000] Command line:
BOOT_IMAGE=/boot/vmlinuz-4.1.0-rc7+
root=UUID=56684438-bf61-422a-9c47-e0d7e405f4e7 ro quiet splash
...
Jul 15 21:20:04 blind-fury kernel: [  173.591327] usbcore: registered new
interface driver btusb
Jul 15 21:20:04 blind-fury kernel: [  173.604148] Bluetooth: hci0: BCM:
chip id 63
Jul 15 21:20:04 blind-fury kernel: [  173.606079] Bluetooth: hci0:
BCM20702A1 (001.002.014) build 0000
Jul 15 21:20:04 blind-fury kernel: [  173.628434] bluetooth hci0: Direct
firmware load for brcm/BCM20702A1-0a5c-21e8.hcd failed with error -2
Jul 15 21:20:04 blind-fury kernel: [  173.628439] Bluetooth: hci0: BCM:
Patch brcm/BCM20702A1-0a5c-21e8.hcd not found
--

These are checked here, but code hasn't changed recently:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/drivers/base/firmware_class.c?id=6593d9245bc66e6e3cf4ba6d365a7833110c1402#n1135


There has been changes to the btbcm.c code wrt to firmware loading:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/bluetooth/btbcm.c?id=18aeb4445aa00f6f402ba3a92a2e9ff3d13882b4

Simon.

Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

[simon]

> It looks like the firmware 'opt_flags' must be different, so this may be a
> contributing factor.

Plot thickens.... kernel config has changed since I built 4.1.0rc7, but I
don't recall doing it or starting a fresh.

/boot/config-4.1.0-rc7+
--
CONFIG_PREVENT_FIRMWARE_BUILD=y
CONFIG_FW_LOADER=y
CONFIG_FIRMWARE_IN_KERNEL=y
CONFIG_EXTRA_FIRMWARE=""
CONFIG_FW_LOADER_USER_HELPER=y
# CONFIG_FW_LOADER_USER_HELPER_FALLBACK is not set
CONFIG_WANT_DEV_COREDUMP=y
CONFIG_ALLOW_DEV_COREDUMP=y
CONFIG_DEV_COREDUMP=y
--

/boot/config-4.2.0-rc1+
--
CONFIG_PREVENT_FIRMWARE_BUILD=y
CONFIG_FW_LOADER=y
CONFIG_FIRMWARE_IN_KERNEL=y
CONFIG_EXTRA_FIRMWARE=""
CONFIG_FW_LOADER_USER_HELPER=y
CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y   <------------!!!
CONFIG_WANT_DEV_COREDUMP=y
CONFIG_ALLOW_DEV_COREDUMP=y
CONFIG_DEV_COREDUMP=y
--


Has a kconfig forced a change?.... Grrr
--
$ git blame ./drivers/leds/Kconfig
--
c93d08fa7 (Milo(Woogyom) Kim         2013-02-05 18:01:23 +0900 228) config
LEDS_LP55XX_COMMON
33b3a561f (Kim, Milo                 2013-07-09 02:11:37 -0700 229)    
tristate "Common Driver for TI/National LP5521/5523/55231/5562/8501"
33b3a561f (Kim, Milo                 2013-07-09 02:11:37 -0700 230)    
depends on LEDS_LP5521 || LEDS_LP5523 || LEDS_LP5562 || LEDS_LP8501
10c06d178 (Milo(Woogyom) Kim         2013-02-05 19:17:20 +0900 231)    
select FW_LOADER
b67893206 (Milo Kim                  2015-06-28 17:39:14 -0700 232)    
select FW_LOADER_USER_HELPER_FALLBACK                                
<-----!!!!
c93d08fa7 (Milo(Woogyom) Kim         2013-02-05 18:01:23 +0900 233)     help
33b3a561f (Kim, Milo                 2013-07-09 02:11:37 -0700 234)      
This option supports common operations for LP5521/5523/55231/5562/8501
c93d08fa7 (Milo(Woogyom) Kim         2013-02-05 18:01:23 +0900 235)      
devices.
--

So in summary this problem is showing up now as the 'User Helper Fallback'
is now forced on, obviously the underlying problem needs to be fixed - but
I don't know when it crept in.

Will try building 4.1 with this option to see if it fails.
Simon

Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

[Kim, Milo]

Hi Simon,

On 7/17/2015 3:14 PM, simon@mungewell.org wrote:
>
>> It looks like the firmware 'opt_flags' must be different, so this may be a
>> contributing factor.
>
> Plot thickens.... kernel config has changed since I built 4.1.0rc7, but I
> don't recall doing it or starting a fresh.
>
> /boot/config-4.1.0-rc7+
> --
> CONFIG_PREVENT_FIRMWARE_BUILD=y
> CONFIG_FW_LOADER=y
> CONFIG_FIRMWARE_IN_KERNEL=y
> CONFIG_EXTRA_FIRMWARE=""
> CONFIG_FW_LOADER_USER_HELPER=y
> # CONFIG_FW_LOADER_USER_HELPER_FALLBACK is not set
> CONFIG_WANT_DEV_COREDUMP=y
> CONFIG_ALLOW_DEV_COREDUMP=y
> CONFIG_DEV_COREDUMP=y
> --
>
> /boot/config-4.2.0-rc1+
> --
> CONFIG_PREVENT_FIRMWARE_BUILD=y
> CONFIG_FW_LOADER=y
> CONFIG_FIRMWARE_IN_KERNEL=y
> CONFIG_EXTRA_FIRMWARE=""
> CONFIG_FW_LOADER_USER_HELPER=y
> CONFIG_FW_LOADER_USER_HELPER_FALLBACK=y   <------------!!!
> CONFIG_WANT_DEV_COREDUMP=y
> CONFIG_ALLOW_DEV_COREDUMP=y
> CONFIG_DEV_COREDUMP=y
> --
>
>
> Has a kconfig forced a change?.... Grrr
> --
> $ git blame ./drivers/leds/Kconfig
> --
> c93d08fa7 (Milo(Woogyom) Kim         2013-02-05 18:01:23 +0900 228) config
> LEDS_LP55XX_COMMON
> 33b3a561f (Kim, Milo                 2013-07-09 02:11:37 -0700 229)
> tristate "Common Driver for TI/National LP5521/5523/55231/5562/8501"
> 33b3a561f (Kim, Milo                 2013-07-09 02:11:37 -0700 230)
> depends on LEDS_LP5521 || LEDS_LP5523 || LEDS_LP5562 || LEDS_LP8501
> 10c06d178 (Milo(Woogyom) Kim         2013-02-05 19:17:20 +0900 231)
> select FW_LOADER
> b67893206 (Milo Kim                  2015-06-28 17:39:14 -0700 232)
> select FW_LOADER_USER_HELPER_FALLBACK
> <-----!!!!
> c93d08fa7 (Milo(Woogyom) Kim         2013-02-05 18:01:23 +0900 233)     help
> 33b3a561f (Kim, Milo                 2013-07-09 02:11:37 -0700 234)
> This option supports common operations for LP5521/5523/55231/5562/8501
> c93d08fa7 (Milo(Woogyom) Kim         2013-02-05 18:01:23 +0900 235)
> devices.
> --
>
> So in summary this problem is showing up now as the 'User Helper Fallback'
> is now forced on, obviously the underlying problem needs to be fixed - but
> I don't know when it crept in.
>

The 'CONFIG_FW_LOADER_USER_HELPER_FALLBACK' enables to load firmware 
data manually by accessing /sys/class/firmware/<name>/data. It runs in 
case the firmware file is missing.
This user helper fallback will be enabled if one of LP55xx driver is 
included in your dot config. Please see my patch below.

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/leds?id=b67893206fc0a0e8af87130e67f3d8ae553fc87c

However, I'm not sure why this affects your system lockup. Can I have 
more details?

Best regards,
Milo

Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

[simon]

>> So in summary this problem is showing up now as the 'User Helper
>> Fallback'
>> is now forced on, obviously the underlying problem needs to be fixed -
>> but
>> I don't know when it crept in.
>>
>
> The 'CONFIG_FW_LOADER_USER_HELPER_FALLBACK' enables to load firmware
> data manually by accessing /sys/class/firmware/<name>/data. It runs in
> case the firmware file is missing.
> This user helper fallback will be enabled if one of LP55xx driver is
> included in your dot config. Please see my patch below.
>
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/leds?id=b67893206fc0a0e8af87130e67f3d8ae553fc87c
>
> However, I'm not sure why this affects your system lockup. Can I have
> more details?

Hi Milo,
I'm not suggesting that your patch is the cause, just that it is an
'enabler' and and explains why the problem (system lockup when I plug USB
Bluetooth dongle in) appears now.

A full Oops log is further back in this thread:
http://www.spinics.net/lists/linux-bluetooth/msg63090.html


>> Will try building 4.1 with this option to see if it fails.

A very quick test as I was leaving the house this morning shows that 4.1
with 'CONFIG_FW_LOADER_USER_HELPER_FALLBACK' does not show the problem.

So at least we know the 'real' problem is a recent change to the code.
Simon

Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

[Laura Abbott]

On 07/17/2015 08:14 AM, simon@mungewell.org wrote:
>
>>> So in summary this problem is showing up now as the 'User Helper
>>> Fallback'
>>> is now forced on, obviously the underlying problem needs to be fixed -
>>> but
>>> I don't know when it crept in.
>>>
>>
>> The 'CONFIG_FW_LOADER_USER_HELPER_FALLBACK' enables to load firmware
>> data manually by accessing /sys/class/firmware/<name>/data. It runs in
>> case the firmware file is missing.
>> This user helper fallback will be enabled if one of LP55xx driver is
>> included in your dot config. Please see my patch below.
>>
>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/leds?id=b67893206fc0a0e8af87130e67f3d8ae553fc87c
>>
>> However, I'm not sure why this affects your system lockup. Can I have
>> more details?
>
> Hi Milo,
> I'm not suggesting that your patch is the cause, just that it is an
> 'enabler' and and explains why the problem (system lockup when I plug USB
> Bluetooth dongle in) appears now.
>
> A full Oops log is further back in this thread:
> http://www.spinics.net/lists/linux-bluetooth/msg63090.html
>
>
>>> Will try building 4.1 with this option to see if it fails.
>
> A very quick test as I was leaving the house this morning shows that 4.1
> with 'CONFIG_FW_LOADER_USER_HELPER_FALLBACK' does not show the problem.
>
> So at least we know the 'real' problem is a recent change to the code.
> Simon
>

I think this was reported and fixed

https://lkml.org/lkml/2015/7/8/858
https://lkml.org/lkml/2015/7/8/1199
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=6f957724b94cb19f5c1c97efd01dd4df8ced323c

Thanks,
Laura

Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

[simon]

> On 07/17/2015 08:14 AM, simon@mungewell.org wrote:
>>
>>>> So in summary this problem is showing up now as the 'User Helper
>>>> Fallback'
>>>> is now forced on, obviously the underlying problem needs to be fixed -
>>>> but
>>>> I don't know when it crept in.
>>>>
>>>
>>> The 'CONFIG_FW_LOADER_USER_HELPER_FALLBACK' enables to load firmware
>>> data manually by accessing /sys/class/firmware/<name>/data. It runs in
>>> case the firmware file is missing.
>>> This user helper fallback will be enabled if one of LP55xx driver is
>>> included in your dot config. Please see my patch below.
>>>
>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/leds?id=b67893206fc0a0e8af87130e67f3d8ae553fc87c
>>>
>>> However, I'm not sure why this affects your system lockup. Can I have
>>> more details?
>>
>> Hi Milo,
>> I'm not suggesting that your patch is the cause, just that it is an
>> 'enabler' and and explains why the problem (system lockup when I plug
>> USB
>> Bluetooth dongle in) appears now.
>>
>> A full Oops log is further back in this thread:
>> http://www.spinics.net/lists/linux-bluetooth/msg63090.html
>>
>>
>>>> Will try building 4.1 with this option to see if it fails.
>>
>> A very quick test as I was leaving the house this morning shows that 4.1
>> with 'CONFIG_FW_LOADER_USER_HELPER_FALLBACK' does not show the problem.
>>
>> So at least we know the 'real' problem is a recent change to the code.
>> Simon
>>
>
> I think this was reported and fixed
>
> https://lkml.org/lkml/2015/7/8/858
> https://lkml.org/lkml/2015/7/8/1199
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=6f957724b94cb19f5c1c97efd01dd4df8ced323c
>

Certainly looks like a plausible solution, will build kernel tonight to
confirm.

If  Shuah is still looking for the trigger, see above note regarding
'CONFIG_FW_LOADER_USER_HELPER_FALLBACK'.

Thanks, and have an awesome weekend. :-)
Simon

Re: Kernel Oops: btusb: 4.2rc1 System lockup with BT dongle insert - log attached

[simon]

>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=6f957724b94cb19f5c1c97efd01dd4df8ced323c
>>
>
> Certainly looks like a plausible solution, will build kernel tonight to
> confirm.

Just to confirm; 4.2rc1 + above patch, and 4.2rc2 both function correctly
and I no longer see the lock up/Oops.

Thanks to all who helped out,
Simon.