Interaction with DS5Dongle on Disconnect

[tb0mb]

Hello,

First, thank you so much for your software. The Impulse Trigger/Adaptive Trigger conversion is so nice. I don't know who would willingly still use DS4Windows!

If you're not familiar with the DS5Dongle project it lets uses a Raspberry Pi to do bluetooth connection to dualsense and present it to the PC as wired connection. Everything is working perfectly with your software and the dongle! Except, when my controller shuts off to save power, PadForge does not "let go" of the wired connection to the dongle. PadForge sees the controller input as still active and actually sets the joysticks to full input, causing problems with regular Windows usage. Interestingly, this behavior does not happen when I use USB. And, if I close PadForge and re-open it, it properly recognizes the controller as having been disconnected.

I recognize this maybe is not a problem with your program, but wondering if you have any suggestions. I am guessing maybe the same way you have the wake deflection setting, the DS5Dongle is doing the same thing, maybe it is detecting an application is still accessing the controller when it shuts off and tries to keep the connection active? Anyways I posted a comment over there.

I tested the same scenario in gamepad testers, they properly let go of the controller when it shuts off and are no longer detected. Windows does still see the dongle as an active controller even when it's off, so that is probably the core issue. Not sure why though for example Steam Input or gamepad tester on web lose connection, but padforge does not.

[hifihedgehog]

The dongle is keeping its DualSense on USB after your controller shuts off, and that part is up to its firmware. A real DualSense drops off the bus when it disconnects, over USB or Bluetooth, and that removal is how SDL, and PadForge through it, know the pad is gone. The dongle uses Sony's DualSense IDs, so while its DualSense interface stays up, Windows and SDL see a connected controller frozen in place. If Keep Controller Awake is on for that slot, PadForge also keeps holding its deflection, because the controller still looks present.

Here's what to pass on to the DS5Dongle maintainer:

In v0.7.2-hotfix, the latest release, the dongle can leave its DualSense HID interface enumerated after the controller disconnects, so hosts keep a connected pad with no controller behind it. A real DualSense leaves the bus at that point, and hosts rely on that removal.

1. With Wake PC from sleep on PS button on, the HCI_EVENT_DISCONNECTION_COMPLETE handler in src/bt.cpp (line 596) never detaches, since it calls tud_disconnect() only when wake is off and the bus isn't suspended (lines 602-603). The interface keeps serving the last report it has (interrupt_loop in src/main.cpp, line 52).
2. With wake off, a controller that disconnects while the bus is suspended is never detached afterward, because tud_resume_cb and tud_mount_cb in src/wake.cpp (lines 170 and 177) don't check for a missing controller.

The master branch already fixes both, in edec7f7 and c67c7f6: with no controller connected, the dongle now drops the DualSense interface on disconnect, on resume and on re-enumeration, keeping only the wake keyboard when wake is on. Neither commit is in a release yet, so a release with them fixes this.

Until there's a release, turning Wake PC from sleep on PS button off, if you have it on, makes v0.7.2 drop the DualSense when the controller shuts off while the PC is awake.

[tb0mb]

Thank you for the detailed response!! I am using the Linux bridge version which has a setting exposed for allow wake, already disabled but not working. I will pass along these comments in hopes it's fixed. I am so happy with this combination already, thank you again! 🙏🏼