I've made some tiny progress on decoding that unrecognized line. Need to make somesort of fake controller and attach that to Windows and trace the usb traffic.

I stole back a F310 Logitech controller from my oldest son (which he stole from me originally). This can be switched between XInput and DirectInput = normal HID.

I will have to live with the badly behaving vusbhci, but now there's a skeleton for xinput.class. It's easier to try interrogate it that way as the hid class won't come in the way.

Has anyone got a Linux setup and some xinput controllers? I'd be interested in the USB HID descriptor, namely those parts of it that Linux lsusb won't parse but only show the content of.

Interface Descriptor:
bLength 9
bDescriptorType 4
bInterfaceNumber 0
bAlternateSetting 0
bNumEndpoints 2
bInterfaceClass 255 Vendor Specific Class
bInterfaceSubClass 93
bInterfaceProtocol 1
iInterface 0
** UNRECOGNIZED: 10 21 10 01 01 24 81 14 03 00 03 13 02 00 03 00

Hi,

I have tried some VM on this Fedora 28 Linux installation, but I like to use the hosted personal build. I don't remember the exact issues, but to use usb devices it's far more easier for me to use the vusbhci and hosted AROS. Tried the native way, no go for higher level coding of the stack, nor lower level...

Chris Hodges has made a stunning USB stack, but unfortunatly it has grown much too large for one person to bear. Poseidon really is truly amazing stack, we only need to bend it a little further... There are some paths taken that aren't really necessary, like the drivers telling what USB versions they support and such, but then binary compatibility steps in...

Some other issues are the use of Forbids to protect the stack and there are a lot of them. Semaphores are the way to go...

Any help is much appreciated. I'm also thinging to drop the never ending pursue of USB3.0 with the current vusbchi ("V"irtual "USB" "H"ost "C"ontroller "I"nterface) on hosted AROS and disectim them to vusbhci2.0 and vusbhci3.0 due to libusb issues and that the driver code would get much too messy.

Any help on the AbortIO and everything else is much appreciated.

Dizzy: Is it the same if you are using some vm-player or is that not what you want?

I can ask Neil.

Quote:

Noticed the AHI debugs, not sure if AHI has ever crashed here on hosted (alsa.device)

My cmdAbort function does not work on vusbhci but hangs the Trident

/home/dzy/AROS_svn/rom/usb/classes/hid/./hid.class.c:hid_usbReleaseInterfaceBinding/339: nepHidReleaseInterfaceBinding(4259eef4)
/home/dzy/AROS_svn/rom/usb/classes/hid/./hid.class.c:hid_nHidTask/1132: Going down the river!
[VUSBHCI] AbortIO: Entering function
[VUSBHCI2.00] cmdAbortIO: Aborting cmdIntXFer ioreq 7ff3a3edb703
[VUSBHCI2.00] cmdAbortIO: Semaphore accuired
[VUSBHCI2.00] cmdAbortIO: 423ed79c == 423ed79c
[VUSBHCI2.00] cmdAbortIO: removed
[VUSBHCI2.00] cmdAbortIO: Returning TRUE

And then Trident just sits there. I've not implemented all the functions (cmdFlush?) so maybe that's why... It will take some time to make vusbhci really usable for implementing some sort of gamepad api

In AROS there is this sound playing when pluging inn and out USB devices. Sometimes I get this sound.datatype crash message but system does not go down.

Hi,

I think I already have similar PS2 to usb converter so no need to send one, but thanks for the offer.

I'm debugging the usb stack on hosted AROS so I first need to get that going a bit better. I already can plug controllers in and they seem to work, but the darn vusbhci.device keeps crashing AROS when I remove and plug again devices or try unbinding classes from the device. Do that many times and Linux logs me out... :) It's scared... :)

Dizzy: Thanks for looking into it.

I did manage to get analogue controllers both left and right to work with Tower57 with a lot of tweaking. I used the analogue hack and cleared all data and filled them inn. This does for sure not work with every controller. There are controllers that give me lot of mess. Some I can not configure or go into at all.
I had conversation through e-mail with the guy that ported Tower57 to AROS and he did this hack for AROS that made L and R analgue controllers work as they should. I had to configure port 1 and 3 for different functions on same controller.

This ps2 adapter.

https://www.ebay.com/itm/PS1-PS2-PSX-Playstation-Dual-Joypad-Game-Controller-to-PC-USB-Converter-Adapter/191947183494?epid=1973745786&hash=item2cb0f19586:g:qekAAOSwmfhX2Pk8

work with normal lowlevel.lib and I can configure buttons, but there are some functions missing. I'm not sure how all the backside of the controller is named but something like this.
L - bumper L - Triger R - bumper L - triger
Select button on Playstation controllers are not named either.

https://www.google.com/search?q=what+are+buttons+on+playstation+controller+called&client=firefox-b&source=lnms&tbm=isch&sa=X&ved=0ahUKEwivv6fZ6cHdAhVD8ywKHQHbD_8Q_AUICigB&biw=1366&bih=654#imgdii=o--AxykSQ_5IxM:&imgrc=JHMIHL_GSOfaTM:

I would like to send you a ps1 or 2 controller with the adapter I linked to over for free.
This is something we could test together and recommend as it seams to be working good.

There should for sure be OS support for analogue controllers like they have in MorphOS and OS4 but that would require a lot of missing stuff.

Other thing I discovered. When I boot up and start a game that only require direction keys and 1 or 2 button. Classic amiga config. everything is good. If I start Tower57 with analogue hack I have to first go into trident.prefs and the controller before it get recognized.
Is there something to add to user-startup or something so this is not needed?
Plug controller inn and out of USB port does not help.

How FPSE use the controller I do not know but it is possible to download and look at the source files form here:

http://www.amidog.se/amiga/bin/FPSE/20151029/FPSE-0.10.6-AX86.tar.gz

This is the 2015 version where controller does not work.

What sort of input methods does the games use? Tower 57, does it poll or use an event based input handler?

What the new gamepad (or what ever it gets named) be? A handler or a library or something else? Should it be able to use something else also than a usb input device?

For a new gamepad.class to Poseidon it would require some hacking as the hid.class claims the usb interface and it is not known until much later that the device is Joystick or Gamepad from the usb descriptor usage tables, so hidd.class would need to broadcast the event's to something and games would connect to that something for their inputs and outputs to the device.

EDIT:

I'm now playing with the hidd.class and integrating the Playstation3 controller to it. The controller returns a bogus usage table so it needs to be replaced with the correct one. This has 17 digital buttons hence the commit to display more than 16 buttons correctly. Also my children have trashed all the original controllers so I now only have a Chinese clone of it which seems to send endpipe full of data and not the requested size as indicated by the usage table parser...

Also it would be nice if the buttons in Trident get named correctly according to the controller.

BSzili: yes, both versions from 2012 work. From 2015 it got broken.

This is what happen in 2015 and I pretty sure what give problems.

151009: I noticed there was an unreleased small update. The only change is in the lowlevel.library joy plugin. It allows two lowlevel ports to be used to emulate one psx pad. I've also put together a little plugin SDK. The sources for a few plugins are included. More plugin sources will be added later when I've fixed some dependencies