HID Mouse Filter driver, developing

[ToS]

Hi all, I have a quick question:

Would anyone (with experience) be interested to write a small upper level mouse filter driver with these features:
- custom setup (like in hidUSBf) so that only certain mice can be selected
- filtering mouse input (preventing it form getting to cursor) and routing it to some sort of IPC channel (I'm open for suggestions)
- each mouse device must have unique ID (simple int or use separate IPC data channel)
- possibly a hi-res timestamp attached to each update (since it seems that in windows, on the application side, not all 1ms has the same length)

That's it mostly. Hope I haven't forgotten anything.
I have tried building this filter with some Windows API combo but that just feelt dirty, unstable and seems like timings vary too much.

[fogmaster]

Hi Tos,
i am interested in your filter driver. It is almost the solution i am looking for. I tried to build a lower level driver using the ddk samples but it does not work yet.
The timestamp is not necessary for my aplication but i need the xy-data of a special mouse to use it in my program. This special device should not be used for the windows mouse cursor. Have you already finished your project? Would you share your sources? Would a cooperation be intersting for you?

Best regards!

[ToS]

I haven't done much with driver filter as I'm having some troubles with initial research (both hardware and software related). I'm starting the work on kernel mode driver filter as soon as possible(I don't believe demo-moufiltr will be used.).

[fogmaster]

you are right. In my opinion a kernel mode driver is the only way to realize it too but in your first post you wrote that you tried to build this filter with some Windows API combo but is unstable. Which functions did you use?

[ToS]

Yes but that method is not time-accurate, I was not able to get the messages in each exact milisecond.

[fogmaster]

but you get the relativ xy-movement and can prevent that the values are used for the windows cursor? That would be great! Especially in user mode program!
I intend to use a trackball for controlling the pan tilt movement of moving lights. I have already written am programm that overwrites some bytes in the process memory of the "grandma onPC" software. Now i need the xy-values of a usb trackball to overwrite the bytes of the simulated trackball in "grandma onPC". Can you help me with your current version of the driver?
Best regards!

[ToS]

Ok, the solution was a combo of RawInput and LL_MouseHook. Each in it's own separate process talking trough a named pipe. RawInput can discern between devices so it deceides which mouse is to be blocked, sends the info to Hook process which is allowed to "cancel" mouse events.
Why Microsoft did not allow RawInput to cancel events, beats me but it's quite stupid.
I'd like to help you with the driver but right now I'm stuck too with KMDF.

[fogmaster]

ok, for now i decided to filter the rawinput according to the device handle. So my programm gets the input only from trackball but the cursor still moved by all mice and trackball. Thats enough for testing but a better solution would be nice. I am looking forward to to read from you Please don't feel scrambled but can you estimate how long it will take to develop your driver?

Best regards

[ToS]

I will know that only when I start working on it (which did not happen yet).
Btw, why aren't you using GlovePIE for your project?

[zestoi]

I will know that only when I start working on it (which did not happen yet).
Btw, why aren't you using GlovePIE for your project?

glovepie does tend to eat quite a lot of cpu - due to the way it loops round the script i guess.

i used it to swallow the input from a usb touchpad and convert to midi. maybe i just made the script too complex.