I have an ultranav usb keyboard, which is recognized by the synaptics_usb driver (model SK-8835, FRU 02R0400, USB device 06CB:0009). When the kernel is built with synaptics_usb support, the device uses the "advanced" non-HID interface and the trackpoint becomes very difficult to use, and certainly does not "feel" like other trackpoint devices. I believe this is due to incorrect interpretation of the data from the relevant USB device; the module comments indeed indicate that the driver has not been well tested with pointing sticks. I want to make some changes to the driver wrt trackpoints, but I wanted to ask for some feedback on this before I go off down the wrong path. Jan Steinhoff kindly directed me here. Anyway, details: Right now, synaptics_usb just sends the raw numbers from the USB responses as input event relative coordinates after trimming off some low-order bits (see synusb_report_stick). That is, we get (0x52, 0x42) from the device, and we report relative movement of around (0x5, 0x4). But the values we get from the USB device don't seem very much like coordinates to me. They appear to be readings from the force sensors on the trackpoint in the x, y, and z directions. From what I have found when reading about trackpoint devices, these forces do not directly correspond to relative cursor movement, but are supposed to be interpreted through a transfer function that translates them to cursor velocity, which is then to be translated into movements at certain intervals. There may be different transfer functions that are possible to use, but one such transfer function I've found is in the trackpoint 4 engineering spec [1]. That's where the various settings like 'sensitivity', 'speed', etc come into play for trackpoint devices where this translation is done in the hardware. That document doesn't give all of the necessary constants for generating that specific function, but it gives a general idea. I've got some code to roughly approximate that curve in an x.org input driver, and it seems to work pretty well considering that I'm making some guesses. But it seems like this translation should occur in the synaptics_usb driver. Or at least, even if it doesn't, the reported values should not be reported as REL_X and REL_Y events, since they are not relative cursor coordinates. The z coordinate still seems like it could just stay as ABS_PRESSURE, and features like "Press to Select" could be implemented in a userspace driver. But I personally don't really care about the z axis at all, since I never use any of the PtS functionality. Furthermore, I really think there should be a runtime option to switch the device between "advanced" mode and HID mode. Especially for the trackpoint, the transfer function is not trivial and its "correctness" may be subjective. Right now a system without a synaptics_usb-like driver is better off than Linux, since such a system would use the generic HID mode, and would be more usable. Maybe even the HID mode should be the default, and we only switch to 'advanced' mode on request, if the user wants to change e.g. sensitivity settings or PtS. So, thoughts? I haven't fiddled with adding anything like this to the synaptics_usb driver yet, but I was planning on doing that as soon as I sent this out and didn't hear any objections. [1] <http://blogs.epfl.ch/icenet/documents/Ykt3Eext.pdf>. A graph of the transfer function is on page 43, and some of the other relevant parameters are described on page 24 and 44. -- Andrew Deason adeason@xxxxxxxx -- To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
