[9fans] Building a Raspberry Pi image / Keyboard support

[9fans] Building a Raspberry Pi image / Keyboard support

[Anant Narayanan]

[-- Attachment #1: Type: text/plain, Size: 1085 bytes --]

It's been fun playing around with Plan 9 on Raspberry Pi - thanks to
everyone who made it happen.

1. I'm trying to get a custom built kernel to boot but not having great
luck, unfortunately. I got as far as "mk 'CONF=pi'" in /sys/src/9/bcm
(those are the latest sources, correct?) which generated a 2M kernel named
's9pi'. I added that to the FAT partition on the standard 9pi SD card image
and edited config.txt to point to it. Now, when I boot I see a rainbow
pattern screen -- switching back to the 9pi kernel in config.txt makes it
boot again.

The kernel I built seems to be a bit different than the one included in the
SD card image, though they are roughly the same size:

$ file s9pi # built kernel
s9pi: Plan 9 executable, ARM 7-something
$ file 9pi # kernel includes in SD card image
9pi: data

What am I doing wrong?

2. A couple of my keyboards don't work with the standard kernel on the SD
card (the mouse works fine). Is /sys/src/omap/usbehciomap.c the right place
for me to start looking into adding support for them?

Thanks again!

-Anant

[-- Attachment #2: Type: text/html, Size: 1239 bytes --]

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Bakul Shah]

On Tue, 01 Jul 2014 23:01:53 PDT Anant Narayanan <anant@kix.in> wrote:
>
> 1. I'm trying to get a custom built kernel to boot but not having great
> luck, unfortunately. I got as far as "mk 'CONF=pi'" in /sys/src/9/bcm
> (those are the latest sources, correct?) which generated a 2M kernel named
> 's9pi'. I added that to the FAT partition on the standard 9pi SD card image
> and edited config.txt to point to it. Now, when I boot I see a rainbow
> pattern screen -- switching back to the 9pi kernel in config.txt makes it
> boot again.

mk should've created 9pi and s9pi.  Copy 9pi to the fat partition, not s9pi.

> What am I doing wrong?

The RPi boot program doesn't know about plan9 executables. It
will just copy the bits from the kernel file specified in
config.txt at address 0x8000 and jump there. See 9/bcm/words
for some details.

> 2. A couple of my keyboards don't work with the standard kernel on the SD
> card (the mouse works fine). Is /sys/src/omap/usbehciomap.c the right place
> for me to start looking into adding support for them?

This thread may help:

https://groups.google.com/forum/#!topic/comp.os.plan9/ycok6NTCWCg

Ideally you shouldn't need more than create a custom kbmap file.

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Anant Narayanan]

[-- Attachment #1: Type: text/plain, Size: 1599 bytes --]

That worked great, thank you!

I was able to get the keyboard to work as well, but it seems there is a
different bug where plugging in both a keyboard and mouse at the same time
causes usb/kb to fail.

-Anant


On Tue, Jul 1, 2014 at 11:32 PM, Bakul Shah <bakul@bitblocks.com> wrote:

> On Tue, 01 Jul 2014 23:01:53 PDT Anant Narayanan <anant@kix.in> wrote:
> >
> > 1. I'm trying to get a custom built kernel to boot but not having great
> > luck, unfortunately. I got as far as "mk 'CONF=pi'" in /sys/src/9/bcm
> > (those are the latest sources, correct?) which generated a 2M kernel
> named
> > 's9pi'. I added that to the FAT partition on the standard 9pi SD card
> image
> > and edited config.txt to point to it. Now, when I boot I see a rainbow
> > pattern screen -- switching back to the 9pi kernel in config.txt makes it
> > boot again.
>
> mk should've created 9pi and s9pi.  Copy 9pi to the fat partition, not
> s9pi.
>
> > What am I doing wrong?
>
> The RPi boot program doesn't know about plan9 executables. It
> will just copy the bits from the kernel file specified in
> config.txt at address 0x8000 and jump there. See 9/bcm/words
> for some details.
>
> > 2. A couple of my keyboards don't work with the standard kernel on the SD
> > card (the mouse works fine). Is /sys/src/omap/usbehciomap.c the right
> place
> > for me to start looking into adding support for them?
>
> This thread may help:
>
> https://groups.google.com/forum/#!topic/comp.os.plan9/ycok6NTCWCg
>
> Ideally you shouldn't need more than create a custom kbmap file.
>
>

[-- Attachment #2: Type: text/html, Size: 2254 bytes --]

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[kokamoto]

> That worked great, thank you!

I'm also considering to buy pi for one of my Plan9 terminal.
In that case I may also have Japanese USB keyboard problem...

What is your feeling of pi as a Plan9 terminal?
such as display size, speed etc.

Kenji

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Steve Simon]

> What is your feeling of pi as a Plan9 terminal?
> such as display size, speed etc.

I use ( am using now) a pi for a plan9 terminal.

it is not very quick, I would not compile on it,
but then that is not what a terminal is for.

If you accept that it is only a terminal then it is superb.
It boots in 2 or 3 seconds and consumes 2 watts.

I use a standard dell keyboard and LCD display at 1920x1600.
I made a special effort to find some IBM/Lenovo 3 button optical
mice which work as you would expect.

its the perfect plan9 terminal IMHO.

-Steve

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Charles Forsyth]

[-- Attachment #1: Type: text/plain, Size: 301 bytes --]

On 5 July 2014 00:45, Anant Narayanan <anant@kix.in> wrote:

> I was able to get the keyboard to work as well, but it seems there is a
> different bug where plugging in both a keyboard and mouse at the same time
> causes usb/kb to fail.


"it works for me" but with one mouse and not another!

[-- Attachment #2: Type: text/html, Size: 590 bytes --]

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Bakul Shah]

[-- Attachment #1: Type: text/plain, Size: 2177 bytes --]

If each individually works, chances are this is power related. Make sure it has enough power (5v @ 2A should more than suffice but not all adapters marked so meet their stated rating). There are two test points on the RPi. The voltage difference between the two should be close to 5V under load. If it has sufficient power, and you can run Linux, you can check out if everything works in Linux. lsusb -v can give you more details including current draw of eac device.

> On Jul 4, 2014, at 4:45 PM, Anant Narayanan <anant@kix.in> wrote:
> 
> That worked great, thank you!
> 
> I was able to get the keyboard to work as well, but it seems there is a different bug where plugging in both a keyboard and mouse at the same time causes usb/kb to fail.
> 
> -Anant
> 
> 
>> On Tue, Jul 1, 2014 at 11:32 PM, Bakul Shah <bakul@bitblocks.com> wrote:
>> On Tue, 01 Jul 2014 23:01:53 PDT Anant Narayanan <anant@kix.in> wrote:
>> >
>> > 1. I'm trying to get a custom built kernel to boot but not having great
>> > luck, unfortunately. I got as far as "mk 'CONF=pi'" in /sys/src/9/bcm
>> > (those are the latest sources, correct?) which generated a 2M kernel named
>> > 's9pi'. I added that to the FAT partition on the standard 9pi SD card image
>> > and edited config.txt to point to it. Now, when I boot I see a rainbow
>> > pattern screen -- switching back to the 9pi kernel in config.txt makes it
>> > boot again.
>> 
>> mk should've created 9pi and s9pi.  Copy 9pi to the fat partition, not s9pi.
>> 
>> > What am I doing wrong?
>> 
>> The RPi boot program doesn't know about plan9 executables. It
>> will just copy the bits from the kernel file specified in
>> config.txt at address 0x8000 and jump there. See 9/bcm/words
>> for some details.
>> 
>> > 2. A couple of my keyboards don't work with the standard kernel on the SD
>> > card (the mouse works fine). Is /sys/src/omap/usbehciomap.c the right place
>> > for me to start looking into adding support for them?
>> 
>> This thread may help:
>> 
>> https://groups.google.com/forum/#!topic/comp.os.plan9/ycok6NTCWCg
>> 
>> Ideally you shouldn't need more than create a custom kbmap file.
> 

[-- Attachment #2: Type: text/html, Size: 2983 bytes --]

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Ramakrishnan Muthukrishnan]

On Sat, Jul 5, 2014 at 5:15 AM, Anant Narayanan <anant@kix.in> wrote:
> That worked great, thank you!
>
> I was able to get the keyboard to work as well, but it seems there is a
> different bug where plugging in both a keyboard and mouse at the same time
> causes usb/kb to fail.

I had a tough time with my kb/mouse [1]. I could "fix" it only by
using a powered hub.

[1] https://www.mail-archive.com/9fans@9fans.net/msg30165.html

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Anant Narayanan]

[-- Attachment #1: Type: text/plain, Size: 975 bytes --]

Thanks for the suggestions. The keyboard + mouse work fine with Linux. In
fact they work on Plan 9 together, but only for a few seconds before I get:

kb: /dev/usb/ep6.1: read: i/o error
kb: exiting
usbotg: ep5.1 error intr 00000082
usb/kb... kb: exiting

Sometimes it recovers, but it's not reliable. It may be related to power,
I'll try to get my hands on a powered hub and report back!

-Anant


On Sat, Jul 5, 2014 at 6:13 AM, Ramakrishnan Muthukrishnan <vu3rdd@gmail.com
> wrote:

> On Sat, Jul 5, 2014 at 5:15 AM, Anant Narayanan <anant@kix.in> wrote:
> > That worked great, thank you!
> >
> > I was able to get the keyboard to work as well, but it seems there is a
> > different bug where plugging in both a keyboard and mouse at the same
> time
> > causes usb/kb to fail.
>
> I had a tough time with my kb/mouse [1]. I could "fix" it only by
> using a powered hub.
>
> [1] https://www.mail-archive.com/9fans@9fans.net/msg30165.html
>
>

[-- Attachment #2: Type: text/html, Size: 1622 bytes --]

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[kokamoto]

> I use a standard dell keyboard and LCD display at 1920x1600.

Wow, you have big screen!
I only have a 20" small LCD TV, I don't watch TV much.

> its the perfect plan9 terminal IMHO.

Indeed!
pi can deal with wireless network?

Kenji

PS: I decided to perchase it.

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Shane Morris]

[-- Attachment #1: Type: text/plain, Size: 919 bytes --]

Kenji, I think others are using ethernet to wifi bridges - if I were in the
same position, I would too. USB wifi support probably isn't what it could
be, for obvious reasons. I was using a DX Chinese wifi bridge until I
realised it was beyond hope, it would shut down within five seconds of
being powered on, and cycle over and over. I even had a Linux Python script
to configure the thing for my old wireless network before I moved...
needless to say, it got thrown in a bin. Hopefully you can do better...!

Good luck!




On Sun, Jul 6, 2014 at 11:15 AM, <kokamoto@hera.eonet.ne.jp> wrote:

> > I use a standard dell keyboard and LCD display at 1920x1600.
>
> Wow, you have big screen!
> I only have a 20" small LCD TV, I don't watch TV much.
>
> > its the perfect plan9 terminal IMHO.
>
> Indeed!
> pi can deal with wireless network?
>
> Kenji
>
> PS: I decided to perchase it.
>
>
>

[-- Attachment #2: Type: text/html, Size: 1361 bytes --]

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[kokamoto]

[-- Attachment #1: Type: text/plain, Size: 905 bytes --]

> PS: I decided to perchase it.

Igot pi board from amazon.jp, and yes it's nice for Plan9 terminal.
I'm using max 1.8A power supply, and usb mose and keyboard.
No problem so far.
I can use Japanese keyboard here.
I'm using 20" Sharp LC-20E90 with HDMI cable, then
I got 1184x624 display size (sigh).
Any idea to exand this?

As I have this TV in a diffrent room from the network rooter,
so I have to use wireless LAN which I've not purchased bridge.
Therefore, this is standalone, isolated Plan9 terminal, Hmmm.

I attached my Japanese keyboard program here, which deals
with multibyte sequence of key codes.  Base is same as Gorka's
(probably) program.   This is the same one which I sent to eric.
If anyone want to modify this for your language, don't warry
about lisense etc.

Just run games/mahjongg, and got good feeling of the graphic
speed from this toy like board!

Kenji

[-- Attachment #2: kbjpn.c --]
[-- Type: text/plain, Size: 19196 bytes --]

/*
 * USB Human Interaction Device: keyboard and mouse.
 *
 * If there's no usb keyboard, it tries to setup the mouse, if any.
 * It should be started at boot time.
 *
 * Mouse events are converted to the format of mouse(3)'s mousein file.
 * Keyboard keycodes are translated to scan codes and sent to kbin(3).
 *
 * If there is no keyboard, it tries to setup the mouse properly, else it falls
 * back to boot protocol.
 */

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

enum
{
	Stoprpt	= -2,
	Tick	= -3,
	Exiting	= -4,

	Msec	= 1000*1000,		/* msec per ns */

	Dwcidle	= 8,
};

typedef struct KDev KDev;
typedef struct Kbd Kbd;
typedef struct Mouse Mouse;
typedef struct Kin Kin;

struct Kbd
{
	Channel*	repeatc;
	Channel*	exitc;
	long		nproc;
};

struct Mouse
{
	int	accel;		/* only for mouse */
};

struct KDev
{
	Dev*	dev;		/* usb device*/
	Dev*	ep;		/* endpoint to get events */
	Kin*	in;		/* used to send events to kernel */
	int	idle;		/* min time between reports (× 4ms) */
	int	bootp;		/* has associated keyboard */
	int	debug;
	Kbd;
	Mouse;
	HidRepTempl templ;
	int	(*ptrvals)(KDev *kd, Chain *ch, int *px, int *py, int *pb);
};

/*
 * Kbdin and mousein files must be shared among all instances.
 */
struct Kin
{
	int	ref;
	int	fd;
	char*	name;
};

/*
 * Map for the logitech bluetooth mouse with 8 buttons and wheels.
 *	{ ptr ->mouse}
 *	{ 0x01, 0x01 },	// left
 *	{ 0x04, 0x02 },	// middle
 *	{ 0x02, 0x04 },	// right
 *	{ 0x40, 0x08 },	// up
 *	{ 0x80, 0x10 },	// down
 *	{ 0x10, 0x08 },	// side up
 *	{ 0x08, 0x10 },	// side down
 *	{ 0x20, 0x02 }, // page
 * besides wheel and regular up/down report the 4th byte as 1/-1
 */

/*
 * key code to scan code; for the page table used by
 * the logitech bluetooth keyboard.  ===deleted by K.Okamoto===
 *
static char sctab[256] =
{
[0x00]	0x0,	0x0,	0x0,	0x0,	0x1e,	0x30,	0x2e,	0x20,
[0x08]	0x12,	0x21,	0x22,	0x23,	0x17,	0x24,	0x25,	0x26,
[0x10]	0x32,	0x31,	0x18,	0x19,	0x10,	0x13,	0x1f,	0x14,
[0x18]	0x16,	0x2f,	0x11,	0x2d,	0x15,	0x2c,	0x2,	0x3,
[0x20]	0x4,	0x5,	0x6,	0x7,	0x8,	0x9,	0xa,	0xb,
[0x28]	0x1c,	0x1,	0xe,	0xf,	0x39,	0xc,	0xd,	0x1a,
[0x30]	0x1b,	0x2b,	0x2b,	0x27,	0x28,	0x29,	0x33,	0x34,
[0x38]	0x35,	0x3a,	0x3b,	0x3c,	0x3d,	0x3e,	0x3f,	0x40,
[0x40]	0x41,	0x42,	0x43,	0x44,	0x57,	0x58,	0x63,	0x46,
[0x48]	0x77,	0x52,	0x47,	0x49,	0x53,	0x4f,	0x51,	0x4d,
[0x50]	0x4b,	0x50,	0x48,	0x45,	0x35,	0x37,	0x4a,	0x4e,
[0x58]	0x1c,	0x4f,	0x50,	0x51,	0x4b,	0x4c,	0x4d,	0x47,
[0x60]	0x48,	0x49,	0x52,	0x53,	0x56,	0x7f,	0x74,	0x75,
[0x68]	0x55,	0x59,	0x5a,	0x5b,	0x5c,	0x5d,	0x5e,	0x5f,
[0x70]	0x78,	0x79,	0x7a,	0x7b,	0x0,	0x0,	0x0,	0x0,
[0x78]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x71,
[0x80]	0x73,	0x72,	0x0,	0x0,	0x0,	0x7c,	0x0,	0x0,
[0x88]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0x90]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0x98]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xa0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xa8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xb0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xb8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xc0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xc8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xd0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xd8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xe0]	0x1d,	0x2a,	0x38,	0x7d,	0x61,	0x36,	0x64,	0x7e,
[0xe8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x73,	0x72,	0x71,
[0xf0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xf8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
};
 */     /* K.Okamoto */

static char sctab[256] = 
{
[0x00]	0x0,	0xff,	0xfc,	0x0,	0x1e,	0x30,	0x2e,	0x20,
[0x08]	0x12,	0x21,	0x22,	0x23,	0x17,	0x24,	0x25,	0x26,
[0x10]	0x32,	0x31,	0x18,	0x19,	0x10,	0x13,	0x1f,	0x14,
[0x18]	0x16,	0x2f,	0x11,	0x2d,	0x15,	0x2c,	0x2,	0x3,
[0x20]	0x4,	0x5,	0x6,	0x7,	0x8,	0x9,	0xa,	0xb,
[0x28]	0x1c,	0x1,	0xe,	0xf,	0x39,	0xc,	0xd,	0x1a,
[0x30]	0x1b,	0x2b,	0x2b,	0x27,	0x28,	0x29,	0x33,	0x34,
[0x38]	0x35,	0x3a,	0x3b,	0x3c,	0x3d,	0x3e,	0x3f,	0x40,
[0x40]	0x41,	0x42,	0x43,	0x44,	0x57,	0x58,	0x37,	0x46,
[0x48]	0xc6,	0x52,	0x47,	0x49,	0x53,	0x4f,	0x51,	0x4d,
[0x50]	0x4b,	0x50,	0x48,	0x45,	0x35,	0x37,	0x4a,	0x4e,
[0x58]	0x1c,	0x4f,	0x50,	0x51,	0x4b,	0x4c,	0x4d,	0x47,
[0x60]	0x48,	0x49,	0x52,	0x53,	0x56,	0x0,	0x5e,	0x59,
[0x68]	0x64,	0x65,	0x66,	0x67,	0x68,	0x69,	0x6a,	0x6b,
[0x70]	0x6c,	0x6d,	0x6e,	0x76,	0x0,	0x0,	0x0,	0x0,
[0x78]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0x80]	0x0,	0x0,	0x0,	0x0,	0x0,	0x7e,	0x0,	0x73,
[0x88]	0x70,	0x7d,	0x79,	0x7b,	0x5c,	0x0,	0x0,	0x0,
[0x90]	0xf2,	0xf1,	0x78,	0x77,	0x76,	0x0,	0x0,	0x0,
[0x98]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xa0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xa8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xb0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xb8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xc0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xc8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xd0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xd8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xe0]	0x1d,	0x2a,	0x38,	0x5b,	0x1d,	0x36,	0x38,	0x5c,
[0xe8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xf0]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
[0xf8]	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,	0x0,
};
static QLock inlck;
static Kin kbdin =
{
	.ref = 0,
	.name = "#Ι/kbin",
	.fd = -1,
};
static Kin ptrin =
{
	.ref = 0,
	.name = "#m/mousein",
	.fd = -1,
};

static int ptrbootpvals(KDev *kd, Chain *ch, int *px, int *py, int *pb);
static int ptrrepvals(KDev *kd, Chain *ch, int *px, int *py, int *pb);

static int
setbootproto(KDev* f, int eid, uchar *, int)
{
	int nr, r, id;

	f->ptrvals = ptrbootpvals;
	r = Rh2d|Rclass|Riface;
	dprint(2, "setting boot protocol\n");
	id = f->dev->usb->ep[eid]->iface->id;
	nr = usbcmd(f->dev, r, Setproto, Bootproto, id, nil, 0);
	if(nr < 0)
		return -1;
	usbcmd(f->dev, r, Setidle, f->idle<<8, id, nil, 0);
	return nr;
}

static uchar ignoredesc[128];

static int
setfirstconfig(KDev* f, int eid, uchar *desc, int descsz)
{
	int nr, r, id, i;

	dprint(2, "setting first config\n");
	if(desc == nil){
		descsz = sizeof ignoredesc;
		desc = ignoredesc;
	}
	id = f->dev->usb->ep[eid]->iface->id;
	r = Rh2d | Rstd | Rdev;
	nr = usbcmd(f->dev,  r, Rsetconf, 1, 0, nil, 0);
	if(nr < 0)
		return -1;
	r = Rh2d | Rclass | Riface;
	nr = usbcmd(f->dev, r, Setidle, f->idle<<8, id, nil, 0);
	if(nr < 0)
		return -1;
	r = Rd2h | Rstd | Riface;
	nr=usbcmd(f->dev,  r, Rgetdesc, Dreport<<8, id, desc, descsz);
	if(nr <= 0)
		return -1;
	if(f->debug){
		fprint(2, "report descriptor:");
		for(i = 0; i < nr; i++){
			if(i%8 == 0)
				fprint(2, "\n\t");
			fprint(2, "%#2.2ux ", desc[i]);
		}
		fprint(2, "\n");
	}
	f->ptrvals = ptrrepvals;
	return nr;
}

/*
 * Try to recover from a babble error. A port reset is the only way out.
 * BUG: we should be careful not to reset a bundle with several devices.
 */
static void
recoverkb(KDev *f)
{
	int i;

	close(f->dev->dfd);		/* it's for usbd now */
	devctl(f->dev, "reset");
	for(i = 0; i < 10; i++){
		if(i == 5)
			f->bootp++;
		sleep(500);
		if(opendevdata(f->dev, ORDWR) >= 0){
			if(f->bootp)
				/* TODO func pointer */
				setbootproto(f, f->ep->id, nil, 0);
			else
				setfirstconfig(f, f->ep->id, nil, 0);
			break;
		}
		/* else usbd still working... */
	}
}

static void
kbfatal(KDev *kd, char *sts)
{
	Dev *dev;

	if(sts != nil)
		fprint(2, "kb: fatal: %s\n", sts);
	else
		fprint(2, "kb: exiting\n");
	if(kd->repeatc != nil){
		chanclose(kd->repeatc);
		for(; kd->nproc != 0; kd->nproc--)
			recvul(kd->exitc);
		chanfree(kd->repeatc);
		chanfree(kd->exitc);
		kd->repeatc = nil;
		kd->exitc = nil;
	}
	dev = kd->dev;
	kd->dev = nil;
	if(kd->ep != nil)
		closedev(kd->ep);
	kd->ep = nil;
	devctl(dev, "detach");
	closedev(dev);
	/*
	 * free(kd); done by closedev.
	 */
	threadexits(sts);
}

static int
scale(KDev *f, int x)
{
	int sign = 1;

	if(x < 0){
		sign = -1;
		x = -x;
	}
	switch(x){
	case 0:
	case 1:
	case 2:
	case 3:
		break;
	case 4:
		x = 6 + (f->accel>>2);
		break;
	case 5:
		x = 9 + (f->accel>>1);
		break;
	default:
		x *= MaxAcc;
		break;
	}
	return sign*x;
}

/*
 * ps2 mouse is processed mostly at interrupt time.
 * for usb we do what we can.
 */
static void
sethipri(void)
{
	char fn[30];
	int fd;

	snprint(fn, sizeof fn, "/proc/%d/ctl", getpid());
	fd = open(fn, OWRITE);
	if(fd >= 0) {
		fprint(fd, "pri 13");
		close(fd);
	}
}

static int
ptrrepvals(KDev *kd, Chain *ch, int *px, int *py, int *pb)
{
	int i, x, y, b, c;
	static char buts[] = {0x0, 0x2, 0x1};

	c = ch->e / 8;

	/* sometimes there is a report id, sometimes not */
	if(c == kd->templ.sz + 1)
		if(ch->buf[0] == kd->templ.id)
			ch->b += 8;
		else
			return -1;
	parsereport(&kd->templ, ch);

	if(kd->debug > 1)
		dumpreport(&kd->templ);
	if(c < 3)
		return -1;
	x = hidifcval(&kd->templ, KindX, 0);
	y = hidifcval(&kd->templ, KindY, 0);
	b = 0;
	for(i = 0; i<sizeof buts; i++)
		b |= (hidifcval(&kd->templ, KindButtons, i) & 1) << buts[i];
	if(c > 3 && hidifcval(&kd->templ, KindWheel, 0) > 0)	/* up */
		b |= 0x10;
	if(c > 3 && hidifcval(&kd->templ, KindWheel, 0) < 0)	/* down */
		b |= 0x08;

	*px = x;
	*py = y;
	*pb = b;
	return 0;
}

static int
ptrbootpvals(KDev *kd, Chain *ch, int *px, int *py, int *pb)
{
	int b, c;
	char x, y;
	static char maptab[] = {0x0, 0x1, 0x4, 0x5, 0x2, 0x3, 0x6, 0x7};

	c = ch->e / 8;
	if(c < 3)
		return -1;
	if(kd->templ.nifcs){
		x = hidifcval(&kd->templ, KindX, 0);
		y = hidifcval(&kd->templ, KindY, 0);
	}else{
		/* no report descriptor for boot protocol */
		x = ((signed char*)ch->buf)[1];
		y = ((signed char*)ch->buf)[2];
	}

	b = maptab[ch->buf[0] & 0x7];
	if(c > 3 && ch->buf[3] == 1)		/* up */
		b |= 0x08;
	if(c > 3 && ch->buf[3] == 0xff)		/* down */
		b |= 0x10;
	*px = x;
	*py = y;
	*pb = b;
	return 0;
}

static void
ptrwork(void* a)
{
	int hipri, mfd, nerrs, x, y, b, c, ptrfd;
	char mbuf[80];
	Chain ch;
	KDev* f = a;

	threadsetname("ptr %s", f->ep->dir);
	hipri = nerrs = 0;
	ptrfd = f->ep->dfd;
	mfd = f->in->fd;
	if(f->ep->maxpkt < 3 || f->ep->maxpkt > MaxChLen)
		kbfatal(f, "weird mouse maxpkt");
	for(;;){
		memset(ch.buf, 0, MaxChLen);
		if(f->ep == nil)
			kbfatal(f, nil);
		c = read(ptrfd, ch.buf, f->ep->maxpkt);
		assert(f->dev != nil);
		assert(f->ep != nil);
		if(c < 0){
			dprint(2, "kb: mouse: %s: read: %r\n", f->ep->dir);
			if(++nerrs < 3){
				recoverkb(f);
				continue;
			}
		}
		if(c <= 0)
			kbfatal(f, nil);
		ch.b = 0;
		ch.e = 8 * c;
		if(f->ptrvals(f, &ch, &x, &y, &b) < 0)
			continue;
		if(f->accel){
			x = scale(f, x);
			y = scale(f, y);
		}
		if(f->debug > 1)
			fprint(2, "kb: m%11d %11d %11d\n", x, y, b);
		seprint(mbuf, mbuf+sizeof(mbuf), "m%11d %11d %11d", x, y,b);
		if(write(mfd, mbuf, strlen(mbuf)) < 0)
			kbfatal(f, "mousein i/o");
		if(hipri == 0){
			sethipri();
			hipri = 1;
		}
	}
}

static void
stoprepeat(KDev *f)
{
	sendul(f->repeatc, Stoprpt);
}

static void
startrepeat(KDev *f, uchar esc1, uchar sc)
{
	ulong c;

	if(esc1)
		c = SCesc1 << 8 | (sc & 0xff);
	else
		c = sc;
	sendul(f->repeatc, c);
}

static void
putscan(KDev *f, uchar esc, uchar sc)
{
	int kbinfd;
	uchar s[2] = {SCesc1, 0};

	kbinfd = f->in->fd;
	if(sc == 0x41){
		f->debug += 2;
		return;
	}
	if(sc == 0x42){
		f->debug = 0;
		return;
	}
	if(f->debug > 1)
		fprint(2, "sc: %x %x\n", (esc? SCesc1: 0), sc);
	s[1] = sc;
	if(esc && sc != 0)
		write(kbinfd, s, 2);
	else if(sc != 0)
		write(kbinfd, s+1, 1);
}

static void
repeattimerproc(void *a)
{
	Channel *c;
	KDev *f;

	threadsetname("kbd reptimer");
	f = a;
	c = f->repeatc;

	for(;;){
		sleep(30);
		if(sendul(c, Tick) == -1)
			break;
	}
	sendul(f->exitc, Exiting);
	threadexits("aborted");
}

static void
repeatproc(void* a)
{
	KDev *f;
	Channel *c;
	int code;
	uint l;
	uvlong timeout;

	threadsetname("kbd repeat");
	f = a;
	c = f->repeatc;

	timeout = 0;
	code = -1;
	for(;;){
		switch(l = recvul(c)){
		default:
			code = l;
			timeout = nsec() + 500*Msec;
			break;
		case ~0ul:
			sendul(f->exitc, Exiting);
			threadexits("aborted");
			break;
		case Stoprpt:
			code = -1;
			break;
		case Tick:
			if(code == -1 || nsec() < timeout)
				continue;
			putscan(f, (uchar)(code>>8), (uchar)code);
			timeout = nsec() + 30*Msec;
			break;
		}
	}
}

static uchar hasesc1(uchar usage)  /* added by K.Okamoto */
{
if(usage==0x46
	 || (usage>0x47 && usage <0x53)
	 || usage==0x54
	 || usage==0x66
	 || ((usage>0xe2 && usage<0xe8) && usage!=0xe5))
	return 1;
else
	return 0;
}


static void
putmod(KDev *f, uchar mods, uchar omods, uchar mask, uchar esc, uchar sc)
{
	/* BUG: Should be a single write */
	if((mods&mask) && !(omods&mask))
		putscan(f, esc, sc);
	if(!(mods&mask) && (omods&mask))
		putscan(f, esc, Keyup|sc);
}

/*
 * This routine diffs the state with the last known state
 * and invents the scan codes that would have been sent
 * by a non-usb keyboard in that case. This also requires supplying
 * the extra esc1 byte as well as keyup flags.
 * The aim is to allow future addition of other keycode pages
 * for other keyboards.
 */
static uchar
putkeys(KDev *f, uchar buf[], uchar obuf[], int n, uchar dk)
{
	int i, j;
	uchar uk;

	putmod(f, buf[0], obuf[0], Mctrl, 0, SCctrl);
	putmod(f, buf[0], obuf[0], (1<<Mlshift), 0, SClshift);
	putmod(f, buf[0], obuf[0], (1<<Mrshift), 0, SCrshift);
	putmod(f, buf[0], obuf[0], Mcompose, 0, SCcompose);
	putmod(f, buf[0], obuf[0], Maltgr, 1, SCcompose);

	/* Report key downs */
	for(i = 2; i < n; i++){
		for(j = 2; j < n; j++)
			if(buf[i] == obuf[j])
			 	break;
		if(j == n && buf[i] != 0){
			dk = sctab[buf[i]];
			putscan(f, hasesc1(dk), dk);
			startrepeat(f, hasesc1(dk), dk);
		}
	}

	/* Report key ups */
	uk = 0;
	for(i = 2; i < n; i++){
		for(j = 2; j < n; j++)
			if(obuf[i] == buf[j])
				break;
		if(j == n && obuf[i] != 0){
			uk = sctab[obuf[i]];
			putscan(f, hasesc1(uk), uk|Keyup);
		}
	}
	if(uk && (dk == 0 || dk == uk)){
		stoprepeat(f);
		dk = 0;
	}
	return dk;
}

static int
kbdbusy(uchar* buf, int n)
{
	int i;

	for(i = 1; i < n; i++)
		if(buf[i] == 0 || buf[i] != buf[0])
			return 0;
	return 1;
}

static void
kbdwork(void *a)
{
	int c, i, kbdfd, nerrs;
	uchar dk, buf[64], lbuf[64];
	char err[128];
	KDev *f = a;

	threadsetname("kbd %s", f->ep->dir);
	kbdfd = f->ep->dfd;

	if(f->ep->maxpkt < 3 || f->ep->maxpkt > sizeof buf)
		kbfatal(f, "weird maxpkt");

	f->exitc = chancreate(sizeof(ulong), 0);
	if(f->exitc == nil)
		kbfatal(f, "chancreate failed");
	f->repeatc = chancreate(sizeof(ulong), 0);
	if(f->repeatc == nil){
		chanfree(f->exitc);
		kbfatal(f, "chancreate failed");
	}

	f->nproc = 2;
	proccreate(repeatproc, f, Stack);
	proccreate(repeattimerproc, f, Stack);
	memset(lbuf, 0, sizeof lbuf);
	dk = nerrs = 0;
	for(;;){
		memset(buf, 0, sizeof buf);
		c = read(kbdfd, buf, f->ep->maxpkt);
		assert(f->dev != nil);
		assert(f->ep != nil);
		if(c < 0){
			rerrstr(err, sizeof(err));
			fprint(2, "kb: %s: read: %s\n", f->ep->dir, err);
			if(strstr(err, "babble") != 0 && ++nerrs < 3){
				recoverkb(f);
				continue;
			}
		}
		if(c <= 0)
			kbfatal(f, nil);
		if(c < 3)
			continue;
		if(kbdbusy(buf + 2, c - 2))
			continue;
		if(usbdebug > 2 || f->debug > 1){
			fprint(2, "kbd mod %x: ", buf[0]);
			for(i = 2; i < c; i++)
				fprint(2, "kc %x ", buf[i]);
			fprint(2, "\n");
		}
		dk = putkeys(f, buf, lbuf, f->ep->maxpkt, dk);
		memmove(lbuf, buf, c);
		nerrs = 0;
	}
}

static void
freekdev(void *a)
{
	KDev *kd;

	kd = a;
	if(kd->in != nil){
		qlock(&inlck);
		if(--kd->in->ref == 0){
			close(kd->in->fd);
			kd->in->fd = -1;
		}
		qunlock(&inlck);
	}
	dprint(2, "freekdev\n");
	free(kd);
}

static void
kbstart(Dev *d, Ep *ep, Kin *in, void (*f)(void*), KDev *kd)
{
	uchar desc[512];
	int n, res;

	qlock(&inlck);
	if(in->fd < 0){
		in->fd = open(in->name, OWRITE);
		if(in->fd < 0){
			fprint(2, "kb: %s: %r\n", in->name);
			qunlock(&inlck);
			return;
		}
	}
	in->ref++;	/* for kd->in = in */
	qunlock(&inlck);
	d->free = freekdev;
	kd->in = in;
	kd->dev = d;
	res = -1;
	kd->ep = openep(d, ep->id);
	if(kd->ep == nil){
		fprint(2, "kb: %s: openep %d: %r\n", d->dir, ep->id);
		return;
	}
	if(in == &kbdin){
		/*
		 * DWC OTG controller misses some split transaction inputs.
		 * Set nonzero idle time to return more frequent reports
		 * of keyboard state, to avoid losing key up/down events.
		 */
		n = read(d->cfd, desc, sizeof desc - 1);
		if(n > 0){
			desc[n] = 0;
			if(strstr((char*)desc, "dwcotg") != nil)
				kd->idle = Dwcidle;
		}
	}
	if(!kd->bootp)
		res= setfirstconfig(kd, ep->id, desc, sizeof desc);
	if(res > 0)
		res = parsereportdesc(&kd->templ, desc, sizeof desc);
	/* if we could not set the first config, we give up */
	if(kd->bootp || res < 0){
		kd->bootp = 1;
		if(setbootproto(kd, ep->id, nil, 0) < 0){
			fprint(2, "kb: %s: bootproto: %r\n", d->dir);
			return;
		}
	}else if(kd->debug)
		dumpreport(&kd->templ);
	if(opendevdata(kd->ep, OREAD) < 0){
		fprint(2, "kb: %s: opendevdata: %r\n", kd->ep->dir);
		closedev(kd->ep);
		kd->ep = nil;
		return;
	}

	incref(d);
	proccreate(f, kd, Stack);
}

static int
usage(void)
{
	werrstr("usage: usb/kb [-bdkm] [-a n] [-N nb]");
	return -1;
}

int
kbmain(Dev *d, int argc, char* argv[])
{
	int bootp, i, kena, pena, accel, devid, debug;
	Ep *ep;
	KDev *kd;
	Usbdev *ud;

	kena = pena = 1;
	bootp = 0;
	accel = 0;
	debug = 0;
	devid = d->id;
	ARGBEGIN{
	case 'a':
		accel = strtol(EARGF(usage()), nil, 0);
		break;
	case 'd':
		debug++;
		break;
	case 'k':
		kena = 1;
		pena = 0;
		break;
	case 'm':
		kena = 0;
		pena = 1;
		break;
	case 'N':
		devid = atoi(EARGF(usage()));		/* ignore dev number */
		break;
	case 'b':
		bootp++;
		break;
	default:
		return usage();
	}ARGEND;
	if(argc != 0)
		return usage();
	USED(devid);
	ud = d->usb;
	d->aux = nil;
	dprint(2, "kb: main: dev %s ref %ld\n", d->dir, d->ref);

	if(kena)
		for(i = 0; i < nelem(ud->ep); i++)
			if((ep = ud->ep[i]) == nil)
				break;
			else if(ep->iface->csp == KbdCSP)
				bootp = 1;

	for(i = 0; i < nelem(ud->ep); i++){
		if((ep = ud->ep[i]) == nil)
			continue;
		if(kena && ep->type == Eintr && ep->dir == Ein &&
		    ep->iface->csp == KbdCSP){
			kd = d->aux = emallocz(sizeof(KDev), 1);
			kd->accel = 0;
			kd->bootp = 1;
			kd->debug = debug;
			kbstart(d, ep, &kbdin, kbdwork, kd);
		}
		if(pena && ep->type == Eintr && ep->dir == Ein &&
		    ep->iface->csp == PtrCSP){
			kd = d->aux = emallocz(sizeof(KDev), 1);
			kd->accel = accel;
			kd->bootp = bootp;
			kd->debug = debug;
			kbstart(d, ep, &ptrin, ptrwork, kd);
		}
	}
	return 0;
}

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[kokamoto]

> I'm using 20" Sharp LC-20E90 with HDMI cable, then
> I got 1184x624 display size (sigh).
> Any idea to exand this?

I got it.
I changes the line of
disable_overscan=1 to uncommented of
the file CONFIG.TXT in /n/9fat directory, and rebooted.

Then, now I have 1280x720 display.

Why you named /dev/sdM0/dos not /dev/sdM0/9fat?
I prefer the latter, because it is accordant with other cpus.

Kenji

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[erik quanstrom]

> I attached my Japanese keyboard program here, which deals
> with multibyte sequence of key codes.  Base is same as Gorka's
> (probably) program.   This is the same one which I sent to eric.
> If anyone want to modify this for your language, don't warry
> about lisense etc.

i think i added this to usb/kb in 9atom.  see the -j option.

- erik

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[erik quanstrom]

On Sat Jul  5 19:46:36 EDT 2014, anant@kix.in wrote:

> Thanks for the suggestions. The keyboard + mouse work fine with Linux. In
> fact they work on Plan 9 together, but only for a few seconds before I get:
> 
> kb: /dev/usb/ep6.1: read: i/o error
> kb: exiting
> usbotg: ep5.1 error intr 00000082
> usb/kb... kb: exiting

i see these sometimes, too.  a formerly working apple full sized wired keyboard
has stopped working with these errors.  i assume it's something tricky.

however, 9atom has some keyboard/mouse fixes that might be worth checking up
on.  the original version asked for pretty big descriptors from the device, and many
devices generate transaction errors, or otherwise do bad things™ when asked for
a descriptor bigger than they envisioned.  the solution is to ask for the minimum size
descriptor plus enough space to get the descriptor length, then ask again with the
device-provided length.

- erik

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[kokamoto]

> i think i added this to usb/kb in 9atom.  see the -j option.

I expected to manage it to use without recompilation for Japanese
user.   Your change still need to recompile the kernel, because
usbd is included in the boot components.
If so, my original version is simpler, I think.
sorry...

Kenji

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[erik quanstrom]

On Mon Jul  7 19:12:32 EDT 2014, kokamoto@hera.eonet.ne.jp wrote:
> > i think i added this to usb/kb in 9atom.  see the -j option.
>
> I expected to manage it to use without recompilation for Japanese
> user.   Your change still need to recompile the kernel, because
> usbd is included in the boot components.
> If so, my original version is simpler, I think.
> sorry...

this is incorrect.

-j is a flag to usb/kb.  this obviously does not require a recompile,
just the addition of kbargs=-j to your plan9.ini.

- erik

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[kokamoto]

> pi can deal with wireless network?

As I have no network connection now.
I applied a USB hub (2.0 3 ports) for use of USB stick.
Yes, it works no problem with usb devices of keyboard+mouse+usb stick.
Please remember I'm using MAX 1.8A power supply.

Kenji

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[kokamoto]

Thanks Richard.

> The convention is that dos is the name for a primary dos partition,
> and 9fat is the name for a dos subpartition at the start of a
> plan9 partition.

Then, I'll follow your decision.

By the way, how I can arrage the correct time in this system?
I don't mean the /adm/timezone/local, because
in the usual PC, we can change it at BIOS screen.

Kenji

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Richard Miller]

> disk/fdisp -p

Oops, disk/fdisk is what I meant to type.

Re: [9fans] Building a Raspberry Pi image / Keyboard support

[Richard Miller]

> Why you named /dev/sdM0/dos not /dev/sdM0/9fat?
> I prefer the latter, because it is accordant with other cpus.

This is standard Plan 9 behaviour of disk/fdisp -p, on any cpu.
The convention is that dos is the name for a primary dos partition,
and 9fat is the name for a dos subpartition at the start of a
plan9 partition.

Perhaps you are thinking of one of the forks which does it
differently?

If you prefer different naming, you can change it yourself with
a 'part' command to /dev/sdM0/ctl - /cfg/$sysname/termrc would
be a sensible place to do this.