cvs commit: src/sys/sys mouse.h src/sys/dev/usb ums.c usbhid.h

[Ian Dowse]

In message <20041212133752.GE68019 at tuatara.fishballoon.org>, Scott Mitchell wri
tes:
>On Sun, Dec 12, 2004 at 12:23:05PM +0000, Ceri Davies wrote:
>> On Sat, Dec 11, 2004 at 10:39:10PM -0700, Scott Long wrote:
>> > Julian Elischer wrote:
>> > >  Add support for USB Microsoft Intellimouse
>> > >  
>> > >  PR:             kern/70607
>> > >  Submitted by:   Matt Wright <matt at consultmatt.co.uk>
>> > >  MFC after:      1 week
>> > 
>> > What 'support' does this add?
>> 
>> They have another axis on the wheel: you can tilt it from left to right.
>> 
>> Ceri
>
>Apart from that, the device in question (wireless Intellimouse 2.0) didn't
>work as any kind of mouse without this patch.
>
>	Scott

As pointed out in the PR though, it doesn't seem necessary to change
struct mousestatus and the MOUSE_GETSTATUS ioctl API just to make
the device work. All other mouse drivers in the tree will need to
be updated to initialise the new `dt' field. There's also one part
of the patch that looks as if it could easily break other devices:

 	ibuf = sc->sc_ibuf;
 	if (sc->sc_iid) {
-		if (*ibuf++ != sc->sc_iid)
-			return;
+		ibuf++;
 	}
 
+	/* The M$ Wireless Intellimouse 2.0 sends 1 extra leading byte of
+	 * data compared to most USB mice. This byte frequently switches
+	 * from 0x01 (usual state) to 0x02. I assume it is to allow
+	 * extra, non-standard, reporting (say battery-life). However
+	 * at the same time it generates a left-click message on the button
+	 * byte which causes spurious left-click's where there shouldn't be.
+	 * This should sort that. */
+	if ((sc->sc_ibuf != ibuf) && (sc->sc_ibuf[0] == 0x02)) 
+		return;
+
 	dx =  hid_get_data(ibuf, &sc->sc_loc_x);
 	dy = -hid_get_data(ibuf, &sc->sc_loc_y);
 	dz = -hid_get_data(ibuf, &sc->sc_loc_z);

BTW, Anish Mistry has been looking at porting a more recent version
of NetBSD's HID fremework to FreeBSD, which I believe may fix the
real bugs that the above code just tries to work around.

Ian