2014-10-22 10:35:13

by Patrick Shirkey

[permalink] [raw]
Subject: rtk_btusb issues

Hi,

I have an android device running the rtk_btusb driver. The original driver
was a couple of years old and there was a problem with stability with BLE.
In addition the localname was not found in the scan_response data.

I upgraded the bluetooth stack by manually backporting from a recent
kernel. That fixed the issue wit the scan_repsonse data but there were
still stability issues with BLE devices (although slightly improved
compared to the older bluetooth stack).

I have tried upgrading the driver to the latest version of the rtk_btusb
driver from the git repo and ported the hooks for bluedroid from the old
driver however I get issues with hci send command when loading the driver.

http://pastebin.com/hXALmXBr

I traced the command but I am not sure where the send value is generated.
I can see the function definition in the hci_dev struct but not the
location where the value is actually set.

I thought it might be an issue specific to bluedroid so I installed bluez
but the issue is still there so it appears to be a bug in my version of
the driver or something wrong with the bluetooth kernel layer.

I went back to the older version of the driver running against bluez
instead of bluedroid. That gets me further along but I still cannot
initialise the bluetooth system on the device.

http://pastebin.com/HcSZXiMu

Does anyone have any suggestions for how to go about resolving this
situation I have got myself into?



--
Patrick Shirkey
Boost Hardware Ltd


2014-10-28 17:27:34

by Marcel Holtmann

[permalink] [raw]
Subject: Re: rtk_btusb issues

Hi Patrik,

>>> I have an android device running the rtk_btusb driver. The original
>>> driver
>>> was a couple of years old and there was a problem with stability with
>>> BLE.
>>> In addition the localname was not found in the scan_response data.
>>>
>>> I upgraded the bluetooth stack by manually backporting from a recent
>>> kernel. That fixed the issue wit the scan_repsonse data but there were
>>> still stability issues with BLE devices (although slightly improved
>>> compared to the older bluetooth stack).
>>>
>>> I have tried upgrading the driver to the latest version of the rtk_btusb
>>> driver from the git repo and ported the hooks for bluedroid from the old
>>> driver however I get issues with hci send command when loading the
>>> driver.
>>>
>>> http://pastebin.com/hXALmXBr
>>>
>>> I traced the command but I am not sure where the send value is
>>> generated.
>>> I can see the function definition in the hci_dev struct but not the
>>> location where the value is actually set.
>>>
>>> I thought it might be an issue specific to bluedroid so I installed
>>> bluez
>>> but the issue is still there so it appears to be a bug in my version of
>>> the driver or something wrong with the bluetooth kernel layer.
>>>
>>> I went back to the older version of the driver running against bluez
>>> instead of bluedroid. That gets me further along but I still cannot
>>> initialise the bluetooth system on the device.
>>>
>>> http://pastebin.com/HcSZXiMu
>>>
>>> Does anyone have any suggestions for how to go about resolving this
>>> situation I have got myself into?
>>
>> the rtk_btusb driver is not an upstream driver and that is the main
>> problem. There has been recently an effort to get btusb support the
>> Realtek devices, but then it went silent again.
>>
>> What you need is having proper Realtek support in btusb and nothing else.
>> Out of tree hacked up vendor drivers are not something any of us will
>> likely have a look at and trying to fix.
>>
>
> My quest is to fix the stability issue not the rtk_btusb driver. I
> spotted the "new" branch in the git tree. I am testing that driver now.

and it would be useful if these patches get upstream.

> Do you know if there are known issues with BLE stability on the 8723au
> chipset?

I have not tested any of the Realtek Bluetooth hardware so far.

Regards

Marcel


2014-10-22 13:42:32

by Patrick Shirkey

[permalink] [raw]
Subject: Re: rtk_btusb issues


On Thu, October 23, 2014 12:05 am, Marcel Holtmann wrote:
> Hi Patrick,
>
>> I have an android device running the rtk_btusb driver. The original
>> driver
>> was a couple of years old and there was a problem with stability with
>> BLE.
>> In addition the localname was not found in the scan_response data.
>>
>> I upgraded the bluetooth stack by manually backporting from a recent
>> kernel. That fixed the issue wit the scan_repsonse data but there were
>> still stability issues with BLE devices (although slightly improved
>> compared to the older bluetooth stack).
>>
>> I have tried upgrading the driver to the latest version of the rtk_btusb
>> driver from the git repo and ported the hooks for bluedroid from the old
>> driver however I get issues with hci send command when loading the
>> driver.
>>
>> http://pastebin.com/hXALmXBr
>>
>> I traced the command but I am not sure where the send value is
>> generated.
>> I can see the function definition in the hci_dev struct but not the
>> location where the value is actually set.
>>
>> I thought it might be an issue specific to bluedroid so I installed
>> bluez
>> but the issue is still there so it appears to be a bug in my version of
>> the driver or something wrong with the bluetooth kernel layer.
>>
>> I went back to the older version of the driver running against bluez
>> instead of bluedroid. That gets me further along but I still cannot
>> initialise the bluetooth system on the device.
>>
>> http://pastebin.com/HcSZXiMu
>>
>> Does anyone have any suggestions for how to go about resolving this
>> situation I have got myself into?
>
> the rtk_btusb driver is not an upstream driver and that is the main
> problem. There has been recently an effort to get btusb support the
> Realtek devices, but then it went silent again.
>
> What you need is having proper Realtek support in btusb and nothing else.
> Out of tree hacked up vendor drivers are not something any of us will
> likely have a look at and trying to fix.
>

My quest is to fix the stability issue not the rtk_btusb driver. I
spotted the "new" branch in the git tree. I am testing that driver now.

Do you know if there are known issues with BLE stability on the 8723au
chipset?




--
Patrick Shirkey
Boost Hardware Ltd

2014-10-22 13:05:45

by Marcel Holtmann

[permalink] [raw]
Subject: Re: rtk_btusb issues

Hi Patrick,

> I have an android device running the rtk_btusb driver. The original driver
> was a couple of years old and there was a problem with stability with BLE.
> In addition the localname was not found in the scan_response data.
>
> I upgraded the bluetooth stack by manually backporting from a recent
> kernel. That fixed the issue wit the scan_repsonse data but there were
> still stability issues with BLE devices (although slightly improved
> compared to the older bluetooth stack).
>
> I have tried upgrading the driver to the latest version of the rtk_btusb
> driver from the git repo and ported the hooks for bluedroid from the old
> driver however I get issues with hci send command when loading the driver.
>
> http://pastebin.com/hXALmXBr
>
> I traced the command but I am not sure where the send value is generated.
> I can see the function definition in the hci_dev struct but not the
> location where the value is actually set.
>
> I thought it might be an issue specific to bluedroid so I installed bluez
> but the issue is still there so it appears to be a bug in my version of
> the driver or something wrong with the bluetooth kernel layer.
>
> I went back to the older version of the driver running against bluez
> instead of bluedroid. That gets me further along but I still cannot
> initialise the bluetooth system on the device.
>
> http://pastebin.com/HcSZXiMu
>
> Does anyone have any suggestions for how to go about resolving this
> situation I have got myself into?

the rtk_btusb driver is not an upstream driver and that is the main problem. There has been recently an effort to get btusb support the Realtek devices, but then it went silent again.

What you need is having proper Realtek support in btusb and nothing else. Out of tree hacked up vendor drivers are not something any of us will likely have a look at and trying to fix.

Regards

Marcel