Middle button on usb mice

Middle button on usb mice

[lemon]

I've recently installed 9front on a lenovo thinkpad x220 and when using
usb mice, the middle button or scroll button (I have a 3 and a 2 button
mouse) doesn't seem to be recognized. The middle button on the laptop
works fine. I've tried ps2, ps2intellimouse and intellimouse with the
laptop trackpad and point both enabled and disabled, and in efi and bios
mode. If I do echo 'debug 2' > /dev/usb/epN.1 then I'll recieve output
onto the console when the middle button is pressed, but cat /dev/mouse
yields nothing. 

The mice behave appropriately under linux and under plan9 on my
raspberry pi. 

What's the next step I can try? I'm very new to plan9 and 9front and I
really appreciate any help you can give me. If this is the wrong mailing list,
please let me know. 

Re: [9front] Middle button on usb mice

[cinap_lenrek]

the mouse port prompt on boot doesnt matter... that applies to
ps/2 and serial mouse only. 

usb mice have no fixed protocol, but a blob (hid descriptor) that
describes the bits in the data packet and how they should be interpreted.

to debug this, first, we need to get rid of the automatically spawned
nusb/kb process thats still attached to the device.

run ps -a | grep kb, you'll get something like:

cinap_lenrek   259856    0:00   0:00      156K Pread    kb [readerproc /dev/usb/ep5.1]

that tells you the assigned device address (5 in this case),
then kill that process:

kill kb | rc

next, we run nusb/kb -ddd 5

replace 5 with the address you got from above there... the
-ddd enables some debug prints. the output will look like this:

report descriptor:
	0x05 0x01 0x09 0x02 0xa1 0x01 0x09 0x01 
	0xa1 0x00 0x05 0x09 0x19 0x01 0x29 0x03 
	0x15 0x00 0x25 0x01 0x95 0x03 0x75 0x01 
	0x81 0x02 0x95 0x01 0x75 0x05 0x81 0x01 
	0x05 0x01 0x09 0x30 0x09 0x31 0x09 0x38 
	0x15 0x81 0x25 0x7f 0x75 0x08 0x95 0x03 
	0x81 0x06 0xc0 0x09 0x3c 0x15 0x00 0x25 
	0x01 0x75 0x01 0x95 0x01 0xb1 0x22 0x95 
	0x07 0xb1 0x01 0xc0 
hidparse: t=8 f=2 usage=90001 v=0
hidparse: t=8 f=2 usage=90002 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=6 usage=10030 v=0
hidparse: t=8 f=6 usage=10031 v=1
hidparse: t=8 f=6 usage=10038 v=0
...

my suspicion is that we eigther fail to read the hid
descriptor (then no "report descriptor:" should be
printed with debug on), and then assume the default
bootproto one for mouse, or the descriptor is malformed
in some way that we do not understand.

the code is here: /sys/src/cmd/nusb/kb/kb.c

the code starts with reading the hid descriptor and
setting the protocol in setproto().

then it spawns readerproc which reads packets from the
endpoint and tries to parse it with repparse(), which
takes the descriptor as input and calls hidparse() for
each item described by the descriptor. hidparse() then
extracts the data from the actual packet for the given
item information.

you can add debug prints if you want, run mk, kill the
old process an start a new one with the usb address
as explained before.

--
cinap

Re: [9front] Middle button on usb mice

[lemon]

On Sat, Sep 24, 2016 at 03:03:23PM +0200, cinap_lenrek@felloff.net wrote:
> my suspicion is that we eigther fail to read the hid
> descriptor (then no "report descriptor:" should be
> printed with debug on), and then assume the default
> bootproto one for mouse, or the descriptor is malformed
> in some way that we do not understand.

running nusb/kb -ddd N returns a "report descripter:" followed by a
bunch of hex similar to your output so the descriptor may be malformed
then?

> the code is here: /sys/src/cmd/nusb/kb/kb.c

I'll see what I can find here. Are there any older versions of this code
I could check out? If there is another version that works, I think it
would be easier to bisect the problem. 

Thanks for your help cinap. If I find a fix, should I add it here or
submit a bug report?

Re: [9front] Middle button on usb mice

[cinap_lenrek]

> running nusb/kb -ddd N returns a "report descripter:" followed by a
> bunch of hex similar to your output so the descriptor may be malformed
> then?

the debug output including the descriptor would be usefull. show what
it outputs when you click the middle mouse button.

> I'll see what I can find here. Are there any older versions of this code
> I could check out?

no.

> If there is another version that works, I think it would be easier to bisect
> the problem. 

no.

--
cinap

Re: [9front] Middle button on usb mice

[lemon]

> the debug output including the descriptor would be usefull. show what
> it outputs when you click the middle mouse button.
report descriptor:
	0x05 0x01 0x09 0x02 0xa1 0x01 0x09 0x01
	0xa1 0x00 0x05 0x09 0x19 0x01 0x29 0x03
	0x15 0x00 0x25 0x01 0x95 0x08 0x75 0x01
	0x81 0x02 0x05 0x01 0x09 0x30 0x09 0x31
	0x09 0x38 0x15 0x81 0x25 0x7f 0x75 0x08
	0x95 0x03 0x81 0x06 0xc0 0xc0
hidparse: t=8 f=2 usage=90001 v=0
hidparse: t=8 f=2 usage=90002 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=6 usage=10030 v=0
hidparse: t=8 f=6 usage=10031 v=0
hidparse: t=8 f=6 usage=10038 v=0

--
lemon

Re: [9front] Middle button on usb mice

[cinap_lenrek]

>> the debug output including the descriptor would be usefull. show what
>> it outputs when you click the middle mouse button.
> report descriptor:
> 	0x05 0x01 0x09 0x02 0xa1 0x01 0x09 0x01
> 	0xa1 0x00 0x05 0x09 0x19 0x01 0x29 0x03
> 	0x15 0x00 0x25 0x01 0x95 0x08 0x75 0x01
> 	0x81 0x02 0x05 0x01 0x09 0x30 0x09 0x31
> 	0x09 0x38 0x15 0x81 0x25 0x7f 0x75 0x08
> 	0x95 0x03 0x81 0x06 0xc0 0xc0

that decodes to:

0x05 0x01 USAGE PAGE(GENERIC DESKTOP)
0x09 0x02 USAGE (Mouse)
0xa1 0x01 COLLECTION (Application)
	0x09 0x01 USAGE (Pointer)
		0xa1 0x00 COLLECTION (Physical)
		0x05 0x09 USAGE PAGE(Button)
		0x19 0x01 USAGE MINIMUM (Button1)
		0x29 0x03 USAGE MAXIMUM (Button3)
		0x15 0x00 LOCAL MINIMUM (0)
		0x25 0x01 LOCAL MAXIMUM (1)
		0x95 0x08 REPORT COUNT (8)
		0x75 0x01 REPORT SIZE (1)
		0x81 0x02 INPUT (Data,Var,Abs)
		0x05 0x01 USAGE
		0x09 0x30 X
		0x09 0x31 Y
		0x09 0x38 WHEEL
		0x15 0x81 Logical min -127
		0x25 0x7f Logical max 127
		0x75 0x08 REPORT SIZE(8)
		0x95 0x03 REPORT COUNT(3)
		0x81 0x06 INPUT (Data,Var,Rel)
	0xc0 END COLLECTION
0xc0 END COLLECTION

note that it maps the buttons 1-3 to bits 0-7 of the first byte;
meaning there are more bits than buttons. normally, the buttons are
mapped like 1-3 -> bits 0-2, and then theres a dummy  report count(1),
report size(5) to cover the rest of the byte.

> hidparse: t=8 f=2 usage=90001 v=0
# button1
> hidparse: t=8 f=2 usage=90002 v=0
# button2

> hidparse: t=8 f=2 usage=90003 v=0
> hidparse: t=8 f=2 usage=90003 v=0
> hidparse: t=8 f=2 usage=90003 v=0
> hidparse: t=8 f=2 usage=90003 v=0
> hidparse: t=8 f=2 usage=90003 v=0
> hidparse: t=8 f=2 usage=90003 v=0
# button3
button3 item is reported multiple times
here (for bits 2-7 of the first byte)

> hidparse: t=8 f=6 usage=10030 v=0
# X
> hidparse: t=8 f=6 usage=10031 v=0
# Y
> hidparse: t=8 f=6 usage=10038 v=0
# Z

this indeed explains why the middle button doesnt work,
but not why the mouse wheel doesnt work... i'll look
in the specification when i get home to see if it sais
anything about disabiguating the max usage vs.
report count case.

anyway, the packet format is 1 byte buttons, and then
3 bytes X,Y,Z in the report packet. you can hexdump
these and veryfy that button3 indeed asserts bit 2 in
the first byte.

--
cinap

Re: [9front] Middle button on usb mice

[cinap_lenrek]

in fact, you might try this (untested):

diff -r 38db1b2cd29e sys/src/cmd/nusb/kb/kb.c
--- a/sys/src/cmd/nusb/kb/kb.c	Thu Sep 22 11:04:43 2016 +0200
+++ b/sys/src/cmd/nusb/kb/kb.c	Wed Sep 28 16:20:13 2016 +0200
@@ -234,8 +234,7 @@
 					for(i=l[UsagMin]; i<=l[UsagMax] && l[UsgCnt] < Nu; i++)
 						l[Nl + l[UsgCnt]++] = i;
 				for(i=0; i<g[RepCnt]; i++){
-					if(i < l[UsgCnt])
-						l[Usage] = l[Nl + i];
+					l[Usage] = i < l[UsgCnt] ? l[Nl + i] : 0;
 					(*f)(t, v, g, l, c, a);
 				}
 				break;

--
cinap

Re: [9front] Middle button on usb mice

[lemon]

> diff -r 38db1b2cd29e sys/src/cmd/nusb/kb/kb.c
> --- a/sys/src/cmd/nusb/kb/kb.c	Thu Sep 22 11:04:43 2016 +0200
> +++ b/sys/src/cmd/nusb/kb/kb.c	Wed Sep 28 16:20:13 2016 +0200
> @@ -234,8 +234,7 @@
>  					for(i=l[UsagMin]; i<=l[UsagMax] && l[UsgCnt] < Nu; i++)
>  						l[Nl + l[UsgCnt]++] = i;
>  				for(i=0; i<g[RepCnt]; i++){
> -					if(i < l[UsgCnt])
> -						l[Usage] = l[Nl + i];
> +					l[Usage] = i < l[UsgCnt] ? l[Nl + i] : 0;
>  					(*f)(t, v, g, l, c, a);
>  				}
>  				break;

Thanks for giving it a shot. It compiles cleanly but did not fix the
issue and when running nusb/kb -ddd N I had the same output as before.
If there's any more information I can share that'll help, let me know.
--
lemon

Re: [9front] Middle button on usb mice

[cinap_lenrek]

> Thanks for giving it a shot. It compiles cleanly but did not fix the
> issue and when running nusb/kb -ddd N I had the same output as before.

are you really sure? you should see:

hidparse: t=8 f=2 usage=90001 v=0
hidparse: t=8 f=2 usage=90002 v=0
hidparse: t=8 f=2 usage=90003 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
...

(notice, usage=90003 should appear only once per report and usage=0
should appear 5 times after the usage=90003 line)

have you run this manually? because to make this effective for hotplug you'll
have to update both /bin/nusb/kb and build a new kernel as the *effective* command
is included in (and run from) the kernel image (so you have keyboard's and mouse
work before a root filesystem can be mounted).

to build new kernel, run mk install in /sys/src/9/pc and then copy
the 9pcf file to 9fat. (9fs 9fat; cp 9pcf /n/9fat)

> If there's any more information I can share that'll help, let me know.

yes, print the first byte from the packet please. like in readerproc()

fprint(2, "p0 = %x\n", p.p[0]);

> --
> lemon

--
cinap

Re: [9front] Middle button on usb mice

[cinap_lenrek]

Attach: /sys/src/cmd/nusb/kb/a.c

i'm pretty confident the change works because i fed your descriptor
into repparse() with the change and i'm getting the expected result:

cpu% 8c -I../lib a.c
cpu% 8l a.8
cpu% ./8.out
hidparse: t=8 f=2 usage=90001 v=0
hidparse: t=8 f=2 usage=90002 v=0
hidparse: t=8 f=2 usage=90003 v=1
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=2 usage=0 v=0
hidparse: t=8 f=6 usage=10030 v=0
hidparse: t=8 f=6 usage=10031 v=0
hidparse: t=8 f=6 usage=10038 v=0

--
cinap

Re: [9front] Middle button on usb mice

[cinap_lenrek]

[-- Attachment #1: Type: text/plain, Size: 33 bytes --]

screwed up attachment..

--
cinap

[-- Attachment #2: a.c --]
[-- Type: text/plain, Size: 5504 bytes --]

#include <u.h>
#include <libc.h>
#include <thread.h>
#include "usb.h"
#include "hid.h"

typedef struct Hidreport Hidreport;
struct Hidreport
{
	int	x;
	int	y;
	int	z;

	int	b;
	int	m;

	int	absx;
	int	absy;
	int	absz;

	int	nk;
	uchar	k[64];

	int	o;
	uchar	*e;
	uchar	p[128];
};

static int debug = 3;

static int
signext(int v, int bits)
{
	int s;

	s = sizeof(v)*8 - bits;
	v <<= s;
	v >>= s;
	return v;
}

static int
getbits(uchar *p, uchar *e, int bits, int off)
{
	int v, m;

	p += off/8;
	off %= 8;
	v = 0;
	m = 1;
	if(p < e){
		while(bits--){
			if(*p & (1<<off))
				v |= m;
			if(++off == 8){
				if(++p >= e)
					break;
				off = 0;
			}
			m <<= 1;
		}
	}
	return v;
}

enum {
	Ng	= RepCnt+1,
	UsgCnt	= Delim+1,	/* fake */
	Nl	= UsgCnt+1,
	Nu	= 256,
};

static uchar*
repparse1(uchar *d, uchar *e, int g[], int l[], int c,
	void (*f)(int t, int v, int g[], int l[], int c, void *a), void *a)
{
	int z, k, t, v, i;

	while(d < e){
		v = 0;
		t = *d++;
		z = t & 3, t >>= 2;
		k = t & 3, t >>= 2;
		switch(z){
		case 3:
			d += 4;
			if(d > e) continue;
			v = d[-4] | d[-3]<<8 | d[-2]<<16 | d[-1]<<24;
			break;
		case 2:
			d += 2;
			if(d > e) continue;
			v = d[-2] | d[-1]<<8;
			break;
		case 1:
			d++;
			if(d > e) continue;
			v = d[-1];
			break;
		}
		switch(k){
		case 0:	/* main item*/
			switch(t){
			case Collection:
				memset(l, 0, Nl*sizeof(l[0]));
				d = repparse1(d, e, g, l, v, f, a);
				continue;
			case CollectionEnd:
				return d;
			case Input:
			case Output:
			case Feature:
				if(l[UsgCnt] == 0 && l[UsagMin] != 0 && l[UsagMin] < l[UsagMax])
					for(i=l[UsagMin]; i<=l[UsagMax] && l[UsgCnt] < Nu; i++)
						l[Nl + l[UsgCnt]++] = i;
				for(i=0; i<g[RepCnt]; i++){
					l[Usage] = i < l[UsgCnt] ? l[Nl + i] : 0;
					(*f)(t, v, g, l, c, a);
				}
				break;
			}
			memset(l, 0, Nl*sizeof(l[0]));
			continue;
		case 1:	/* global item */
			if(t == Push){
				int w[Ng];
				memmove(w, g, sizeof(w));
				d = repparse1(d, e, w, l, c, f, a);
			} else if(t == Pop){
				return d;
			} else if(t < Ng){
				if(t == RepId)
					v &= 0xFF;
				else if(t == UsagPg)
					v &= 0xFFFF;
				else if(t != RepSize && t != RepCnt){
					v = signext(v, (z == 3) ? 32 : 8*z);
				}
				g[t] = v;
			}
			continue;
		case 2:	/* local item */
			if(l[Delim] != 0)
				continue;
			if(t == Delim){
				l[Delim] = 1;
			} else if(t < Delim){
				if(z != 3 && (t == Usage || t == UsagMin || t == UsagMax))
					v = (v & 0xFFFF) | (g[UsagPg] << 16);
				l[t] = v;
				if(t == Usage && l[UsgCnt] < Nu)
					l[Nl + l[UsgCnt]++] = v;
			}
			continue;
		case 3:	/* long item */
			if(t == 15)
				d += v & 0xFF;
			continue;
		}
	}
	return d;
}

/*
 * parse the report descriptor and call f for every (Input, Output
 * and Feature) main item as often as it would appear in the report
 * data packet.
 */
static void
repparse(uchar *d, uchar *e,
	void (*f)(int t, int v, int g[], int l[], int c, void *a), void *a)
{
	int l[Nl+Nu], g[Ng];

	memset(l, 0, sizeof(l));
	memset(g, 0, sizeof(g));
	repparse1(d, e, g, l, 0, f, a);
}

static void
hidparse(int t, int f, int g[], int l[], int, void *a)
{
	Hidreport *p = a;
	int v, m;

	if(t != Input)
		return;
	if(g[RepId] != 0){
		if(p->p[0] != g[RepId]){
			p->o = 0;
			return;
		}
		if(p->o < 8)
			p->o = 8;	/* skip report id byte */
	}
	v = getbits(p->p, p->e, g[RepSize], p->o);
	p->o += g[RepSize];

	if((f & (Fconst|Fdata)) != Fdata)
		return;

	if(debug > 1)
		fprint(2, "hidparse: t=%x f=%x usage=%x v=%x\n", t, f, l[Usage], v);

	if((l[Usage]>>16) == 0x07){	/* keycode */
		if((f & (Fvar|Farray)) == Fvar)
			if(v != 0) v = l[Usage] & 0xFF;
		if(p->nk < nelem(p->k) && v != 0)
			p->k[p->nk++] = v;
		return;
	}

	if(g[LogiMin] < 0)
		v = signext(v, g[RepSize]);
	if((f & (Fvar|Farray)) == Fvar && v >= g[LogiMin] && v <= g[LogiMax]){
		/*
		 * we use logical units below, but need the
		 * sign to be correct for mouse deltas.
		 * so if physical unit is signed but logical
		 * is unsigned, convert to signed but in logical
		 * units.
		 */
		if((f & (Fabs|Frel)) == Frel
		&& g[PhysMin] < 0 && g[PhysMax] > 0
		&& g[LogiMin] >= 0 && g[LogiMin] < g[LogiMax])
			v -= (g[PhysMax] * (g[LogiMax] - g[LogiMin])) / (g[PhysMax] - g[PhysMin]);

		switch(l[Usage]){
		case 0x090001:
		case 0x090002:
		case 0x090003:
		case 0x090004:
		case 0x090005:
		case 0x090006:
		case 0x090007:
		case 0x090008:
			m = 1<<(l[Usage] - 0x090001);
			p->m |= m;
			p->b &= ~m;
			if(v != 0)
				p->b |= m;
			break;
		case 0x010030:
			if((f & (Fabs|Frel)) == Fabs){
				p->x = (v - p->absx);
				p->absx = v;
			} else {
				p->x = v;
				p->absx += v;
			}
			break;
		case 0x010031:
			if((f & (Fabs|Frel)) == Fabs){
				p->y = (v - p->absy);
				p->absy = v;
			} else {
				p->y = v;
				p->absy += v;
			}
			break;
		case 0x010038:
			if((f & (Fabs|Frel)) == Fabs){
				p->z = (v - p->absz);
				p->absz = v;
			} else {
				p->z = v;
				p->absz += v;
			}
			break;
		}
	}
}

static void
usage(void)
{
	fprint(2, "usage: %s [-d] devid\n", argv0);
	threadexits("usage");
}

static uchar rep[] ={
	0x05, 0x01, 0x09, 0x02, 0xa1, 0x01, 0x09, 0x01,
	0xa1, 0x00, 0x05, 0x09, 0x19, 0x01, 0x29, 0x03,
	0x15, 0x00, 0x25, 0x01, 0x95, 0x08, 0x75, 0x01,
	0x81, 0x02, 0x05, 0x01, 0x09, 0x30, 0x09, 0x31,
	0x09, 0x38, 0x15, 0x81, 0x25, 0x7f, 0x75, 0x08,
	0x95, 0x03, 0x81, 0x06, 0xc0, 0xc0,
};

void
threadmain(int argc, char* argv[])
{
	Hidreport p = {0};

	p.o = 0;
	p.e = p.p + 3;
	p.p[0] = 1<<2;
	repparse(rep, rep+sizeof(rep), hidparse, &p);
}

Re: [9front] Middle button on usb mice

[lemon]

On Fri, Sep 30, 2016 at 08:41:16AM +0200, cinap_lenrek@felloff.net wrote:
> update both /bin/nusb/kb and build a new kernel 
Just to follow up, (a bit late I know), but your provided fix worked
perfectly, thank you.

--
lemon