Sony DualShock 3 - call for testers: "sixaxis" script module

psyke83

@sano said in Sony DualShock 3 - call for testers: "sixaxis" script module:

Same behavior : No unwanted input with USB, a lot through BT :(

Bizarre. Do you have other Bluetooth peripherals paired during testing? Perhaps if you explicitly remove everything but the controller from the stack it may resolve the issue.

Do you use the onboard wifi? Perhaps if you temporarily disconnect while testing the
controller, it can help to identify if we're seeing Bluetooth coexistence issues.

I'm using the onboard Bluetooth on my Pi3 and have never seen issues such as this - but I use ethernet instead of wifi, so that may be explain the discrepancy.

Sano

I only had one controller connected at a time.
Moreover, I use ethernet, and disabled wifi via /boot/config.txt.
The only other BT device in range is the my wife's Fitbit Flex, but it's obviously not paired with the Pi.
Guess I'll need to find an external BT adapter and do more tests :)

psyke83

@sano said in Sony DualShock 3 - call for testers: "sixaxis" script module:

I only had one controller connected at a time.
Moreover, I use ethernet, and disabled wifi via /boot/config.txt.
The only other BT device in range is the my wife's Fitbit Flex, but it's obviously not paired with the Pi.
Guess I'll need to find an external BT adapter and do more tests :)

Well, do keep me updated on any developments.

I just disabled wifi via dtoverlay=pi3-disable-wifi, and still can't reproduce the issue you're seeing. Maybe as a last resort, you could try to reset your controller using the pinhole at the back of the controller and then re-pair, but I doubt it'll help.

You never had issues with ps3controller causing erratic axis events like this?

psyke83

@Sano,

If you're up for a possible alternate to using a USB BT adapter, I have a suggestion:

Edit /usr/bin/btuartas superuser and change the following line:

                $HCIATTACH /dev/serial1 bcm43xx 921600 noflow - $BDADDR

to:

                $HCIATTACH /dev/serial1 bcm43xx 460800 noflow - $BDADDR

Then reboot your Pi and see if Bluetooth works more reliably.

Also, can you let me know the Pi revision you have?

cat /proc/cpuinfo

Although I don't experience the same issue as you, I do notice a lot of Bluetooth: hci0: Frame reassembly failed (-84)messages, which led me to this discussion: https://github.com/raspberrypi/linux/issues/2264

My Pi 3 is a revision 1.2 board (a02082).

Thanks.

Sano

@psyke83

Thanks for the tip.
Actually I've done a lot of research and found the same page as you (amongst other), that's why I was looking to test with an external adapter.

Here is my RPi revision from cpuinfo (the exact same as yours BTW) :
Revision : a02082

I followed your advice and reduced the connection speed, and the hci(0) don't appear anymore. Moreover, no more unwanted input on analog sticks.
Nice shot !

Do you plan to merge your branch with master soon ?

Edit : actually there are still some hci errors, but very few compared to before the speed change.

psyke83

Unfortunately, it's going to take some time before you'll see this merged in any form as we're currently busy testing everything for stretch compatibility.

Additionally, there are issues with this scriptmodule that are going to require some evaluation:

I've backported the hid-sony driver from the 4.15 kernel, but patched it to revert to the old button mappings that were used by the driver in the 4.9 kernel (as well as the ps3controller driver), with the intention of minimizing user frustration from getting stuck in menus/having to remap their controller, after installing this module. I'm not sure this is a good idea, because the next Raspberry Pi firmware update on stretch will bump the kernel to 4.14.x, and the hid-sony driver from this kernel release will start using the newer button mappings anyway.
In order to support third-party (Shanwan/Gasia) controllers, we need BlueZ 5.48. The packages I've provided work well, but there are always potential pitfalls, complications and upgrade issues when overwriting distribution versions of packages. This won't be an issue once the official Debian/Raspbian packages are upgraded to this version, but the question is how long that will take to happen, or if it will happen at all in the case of stretch.
I haven't yet received any feedback on third-party Gasia controller compatibility. I'm fairly sure it will work just as well as Shanwan, but I need to be sure in order for us to consider this a full replacement for the old ps3controller driver.

Sano

@psyke83
I see.
I saw the future 4.14 kernel/firmware upgrade as stated above through rpi-update github.

Regarding button mapping, IMHO, I think it would be better to stay as close as possible of the upstream packages/drivers.
Moreover, the stretch image is still a beta and apparently no direct upgrade path from jessie will be supported, so we could hope than by integrating this directly in stretch, very few users will be bothered by controller remapping, because the new installs will use the new mapping anyway...

My two cents.

Oh, and thanks again for your work and advices !

psyke83

@Sano,

Yes, but stretch is not yet using the new mappings. Once the next firmware package is released, I'd certainly be happy to drop the old mappings.

BTW, one more thing you could try is to verify that it's not BlueZ 5.48 causing your issue with Bluetooth stability.

Revert the change to /usr/bin/btuart
Revert the Bluetooth packages back to Raspbian defaults:
sudo apt-get install $(dpkg -l | grep 5.48-1~rpi1 | awk -v os_rel=$(lsb_release -sc) '{print $2"/"os_rel}')

Reboot and see if your controller is stable with the default BlueZ packages. To go back to the new BlueZ packages, just re-install the sixaxis module.

Sano

@psyke83
Just tested as requested.
hci0: Frame reassembly failed (-84) errors are back, as well as unwanted analog stick events.
Going back to the working solution then :)

psyke83

@sano said in Sony DualShock 3 - call for testers: "sixaxis" script module:

@psyke83
Just tested as requested.
hci0: Frame reassembly failed (-84) errors are back, as well as unwanted analog stick events.
Going back to the working solution then :)

At least we know it's not caused by my custom BlueZ packages - much appreciated! I'm still curious as to why I don't experience the same issue, given that we have the exact same revision of Pi 3. Hopefully more people will test this to help discern if it's going to be a common issue.

Sano

@psyke83
Could the answer be that my RPi is near my wifi router (maybe 30cm) ?
Nevertheless those unwanted stick inputs were not there a couple of weeks ago.
Maybe a new firmware/driver/config rendering the BT controller more sensitive ?

psyke83

@sano said in Sony DualShock 3 - call for testers: "sixaxis" script module:

@psyke83
Could the answer be that my RPi is near my wifi router (maybe 30cm) ?
Nevertheless those unwanted stick inputs were not there a couple of weeks ago.
Maybe a new firmware/driver/config rendering the BT controller more sensitive ?

Could be, or perhaps some interference if you have your Pi inside a metal case. Are you on jessie or the stretch testing image? The latter may have different BT/wifi firmware that can affect how Bluetooth behaves.

Sano

@psyke83
My Pi is in a plastic case, and I use the stretch image for a few weeks.
Still trying to find bugs with my "regular" use of retropie, but no big deal ATM.

pjft

I have seen those hci0: Frame reassembly failed (-84) errors in the past in my setup, before I even tried to play with these new drivers. I never got to the bottom of it, though since it didn't seem to have any material impact on the performance I just ignored them.

Sano

@pjft
Just FYI those errors are logged in /var/log/kern.log*.
On my setup that was several MB of logs rewritten on a daily basis (I tested and had 500kB of logs for a few minutes of DS3 use), in addition to the random inputs.
Even with a F2FS formatted root partition, I don't like having error logs writing over and over a a several MB per day on a SD card for nothing :)
Gasp, my OCD is still kicking here...

psyke83

@pjft said in Sony DualShock 3 - call for testers: "sixaxis" script module:

I have seen those hci0: Frame reassembly failed (-84) errors in the past in my setup, before I even tried to play with these new drivers. I never got to the bottom of it, though since it didn't seem to have any material impact on the performance I just ignored them.

Yeah, those messages aren't unique to the BlueZ sixaxis plugin - they definitely occurred with sixad (ps3controller) too. I don't recall seeing complaints about ghost analog movements, though.

It's hard to debug sixad because the virtual uinput device reports a lot of accelerometer events, and doesn't support setting axis fuzz (needed to skip reporting events when there's a small deviation).

psyke83

@sano said in Sony DualShock 3 - call for testers: "sixaxis" script module:

@pjft
Just FYI those errors are logged in /var/log/kern.log*.
On my setup that was several MB of logs rewritten on a daily basis (I tested and had 500kB of logs for a few minutes of DS3 use), in addition to the random inputs.
Even with a F2FS formatted root partition, I don't like having error logs writing over and over a a several MB per day on a SD card for nothing :)
Gasp, my OCD is still kicking here...

That seems quite excessive. I would see perhaps a dozen or so messages in one day at most, not several megabytes worth. I haven't seen any warnings at all since reducing the UART speed.

Perhaps it's due to overclocking your Pi?

Sano

@psyke83
No overclock :)
I have a personal motto to stay as vanilla as possible.
The only changes I make on my installs are :

disabling overscan
disabling wifi
migrate root partition to F2FS

Here is an example of log I had for a single second before reducing speed :

Feb 24 09:51:37 retropie kernel: [552204.012012] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.012040] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.012088] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.012184] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.012274] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.012353] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.114573] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.114577] Bluetooth: hci0 ACL packet for unknown connection handle 520
Feb 24 09:51:37 retropie kernel: [552204.114642] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.114728] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.114813] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.114900] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.114966] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.181939] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.181991] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.182071] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.182165] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.182252] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.182340] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.182426] Bluetooth: hci0: Frame reassembly failed (-84)
Feb 24 09:51:37 retropie kernel: [552204.182480] Bluetooth: hci0: Frame reassembly failed (-84)

psyke83

@sano said in Sony DualShock 3 - call for testers: "sixaxis" script module:

@psyke83
No overclock :)
I have a personal motto to stay as vanilla as possible.

It's good that we have a mitigation for the issue; regardless, if your warranty isn't void, perhaps you could report the issue to the firmware issues tracker. No need to go into detail about RetroPie or this scriptmodule, as you can replicate the behaviour with stock bluez and the default hid_sony driver on the current stretch kernel.

I have a suspicion that your Pi's bluetooth module may be faulty, considering that we have the same revision of Pi. I've often seen them offer to replace boards when a HW fault is diagnosed (but of course, I'm only speculating about your board).

To be clear, the problem is not that the frame reassembly errors are appearing; our revision board doesn't support HW flow control, so those messages are (unfortunately) expected. The reason I suspect a HW fault is due to the magnitude to which the error is occurring for you along with the phantom feedback. Perhaps if you can test in a "quieter" environment and rule out interference etc., it might help. I'd like to see more people test this module so we can get a wider assessment as to whether this would be a common problem.

MadHorse

Hi all,
i have a Shanwan and a Gaisa controller, i would like to try sixaxis.
I installed sixaxis and automatically it will install customhidsony.
Becasue my controllers are not sony i try to install custombluez but it give me an error because a lot of package are missing

dh-systemd git-buildpackage icu-devtools libcap-ng-dev libdbus-glib-1-dev libdw-dev libical-dev libicu-dev libreadline-dev libsubunit-dev libsubunit0 libtinfo-dev libical2

Tried to install all the packages manually with the command

sudo apt-get update
sudo apt-get install dh-systemd git-buildpackage icu-devtools libcap-ng-dev libdbus-glib-1-dev libdw-dev libical-dev libicu-dev libreadline-dev libsubunit-dev libsubunit0 libtinfo-dev libical2

I was able to install all the packages except libical2
E: Unable to locate package libical2.
Someone can help me ?
I'm using Retropie 4.3.12 on a Raspberry Pi 3.

Thank you