linuxconsole-dev

Re: I'm going mad: /dev/tty - device not configured!

[James S. <jsi...@ac...>]

On Thu, 21 Sep 2000, Paolo Scaffardi wrote:

> Please, could you explain me differences between /dev/tty, /dev/tty0 and
> /dev/console?

/dev/console. 
    This is the system console. What display the kernel messages get
    sent to. 

/dev/tty 
    TTY the current process is attached to. Could be a VT, serial console,
    or even a modem.

/dev/tty0
    The current video terminal that is being displayed.

> I have an embedded linux box with a custom keyboard and a custom frame
> buffer. The keyboard driver sends scancodes to 'handle_scancode' in
> keyboard.c and it is not registered as console, or any type of device.

To get your keyboard to work with the console you need a function like

void __init xxxkbd_init_hw(void) {
}

Where xxx is what you want to call your keybaord driver. Take a look at
pckbd_init_hw. See in tty_io.c tty_init is called which calls kbd_init
which in turn calls kbd_init_hw which calls the specific xxxkbd_init_hw
function. This sets up your keyboard for the console system. Of course
you will need a little more than this to get the keyboard working right
but these are the key parts you need. Understand also the standard tree
only works with one keyboard at a time.
 
> The only problem is that when using other programs, like GII, that opens
> /dev/tty to access the keyboard, they dont work and report:
> 
>     /dev/tty - device not configured
>
> Who has to register /dev/tty? How could my keyboard be reached by /dev/tty?
> 
> I'm going mad... 8-(((

See above.

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Re: Console over the video on my frame-buffer device

[James S. <jsi...@ac...>]

On Thu, 21 Sep 2000, Paolo Scaffardi wrote:

> Using graphic applications based on framebuffers (GGI, for example) the
> console is not hidden when a graphic mode is set. This happens on my
> framebuffer device, but i dont know who has to disable the console while
> using a graphi mode.

No this is normal. Remember fbcon is a graphical console. Since /dev/fb is
used to change the video mode we might want to preserve what on the
console. Now if you open /dev/fb to use it as a graphical interface then
you need to mmap the video memory and clear this memory.

This is strange. For most framebuffer based video hardware when you reset
the video mode it clears the video mode.

Re: Custom serial keyboard: playing dirty with the serial console

[James S. <jsi...@ac...>]

> But 'handle_scancode' only prints characters to the console, without 
> apparently processing them. I see everything i write, but when i press
> enter (scancode 0x1c), i only see a new line in the console, not any new
> bash prompt.

Hum? Try running the program showkey. It prints out the key codes for the
keys you press. Can you post the results. If you are translating the codes
into XT scancodes then it shpould make what I get with showkey. BTW make
sure  you tell me what keys you pressed.


P.S.
 
  As I move to californis these next 2 weeks my email service will be
switching. This means I will not be able to read my email from  Friday
until October 2. 

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Console over the video on my frame-buffer device

[Paolo S. <ar...@ti...>]

Using graphic applications based on framebuffers (GGI, for example) the
console is not hidden when a graphic mode is set. This happens on my
framebuffer device, but i dont know who has to disable the console while
using a graphi mode.

I think that framebuffer devices on PowerPC are different from PC, 'cause
with PC, text-mode and graph-mode memories were not the same, as here is.

Do you know anything helpfull for me?

thanx again!

- may the pinguin b with u -

Have a nice day,
Paolo Scaffardi.
AIRVENT SAM s.p.a. - RIMINI - ITALY

I'm going mad: /dev/tty - device not configured!

[Paolo S. <ar...@ti...>]

Please, could you explain me differences between /dev/tty, /dev/tty0 and
/dev/console?

I have an embedded linux box with a custom keyboard and a custom frame
buffer. The keyboard driver sends scancodes to 'handle_scancode' in
keyboard.c and it is not registered as console, or any type of device.

At the boot i dont specify anything to boot the kernel.

It uses correctly both the frame buffer console and the keyboard correctly.

The only problem is that when using other programs, like GII, that opens
/dev/tty to access the keyboard, they dont work and report:

    /dev/tty - device not configured

Who has to register /dev/tty? How could my keyboard be reached by /dev/tty?

I'm going mad... 8-(((

- may the pinguin b with u -

Have a nice day,
Paolo Scaffardi.
AIRVENT SAM s.p.a. - RIMINI - ITALY

Custom serial keyboard: playing dirty with the serial console

[Paolo S. <ar...@ti...>]

We have a custom IR keyboard & mouse, that i have attached to the serial =
port of my board (i have an embedded powerpc). Its protocol is called =
4PPM, and i had to write a driver that intercepts these serial =
characters, decodes and convert them to standard XT scancodes.

Serial routines were implemented, yet, and i decided to 'insert' my 4PPM =
layer inside of the serial console driver.=20

Everything is now ok: i receive correct packets, convert them to XT =
scancodes and call 'handle_scancode' to process them.

But 'handle_scancode' only prints characters to the console, without =
apparently processing them. I see everything i write, but when i press =
enter (scancode 0x1c), i only see a new line in the console, not any new =
bash prompt.

Why this? Is it correct to intercept the serial console and not =
returning single characters from inside it but from 'handle_scancode' =
mode?

Who know that?

THANX!

- may the pinguin b with u -

Have a nice day,
Paolo Scaffardi.
AIRVENT SAM s.p.a. - RIMINI - ITALY

Re: New VGA console coming

[Petr V. <VAN...@vc...>]

On 18 Sep 00 at 10:22, Steffen Seeger wrote:
> 
> However, I don't see this need, as with about 500 Glyphs you can cover most
> of the european languages (german, greek, ...) sufficiently. If you need
> console support for Eastern (e.g. except Korean) languages, you are better of
> to not implement this inside the kernel. Mainly because input requires a
> dictionary search sometimes.

Please, no! I'm trying to write administrator utility for NDS and I want
to support all 0x110000 characters possible... It is true that Novell's 
NWADMIN/NETADMIN/ConsoleOne handle only small portion (normal versions
256 characters from your 8bit codepage), but I want to be better...

Of course, if you have equivalent userspace (library) solution, fine...
But it must be completely seamless, so that 'nwdsread | tee output' still
produces correct japan glyphs on screen.

                                        Best regards,
                                                Petr Vandrovec
                                                van...@vc...
                                                

Re: New VGA console coming

[Steffen S. <se...@ph...>]

James Simmons wrote:
> 
> Sorry I have been away but I was sick.
> 
> > The console layer handles the glyph information to be displayed using full
> > UNICODE-2.0 (or it's successor, the ISO norm for 32bit character codes).
> 
> Same here. By the way I have expanded the conmakehash.c program to
> generate a unicode->font map of up to 65,536 characters.
>
> > Using the font and attribute information, this is translated into a
> > font-specific code, which is stored in the shadow buffer using a pattern
> > defined by the graphics driver only. Then the graphics driver does the
> > rendering from the shadow buffer, it never gets involved with the
> > glyph-code -> font position translation. This way it is up to the graphics
> > driver, which attributes and how many font characters it supports.
> 
> So the shadow buffer is driver specific. Couldn't the shadow buffer get
> really huge for some types of hardware ? 

No, as the shadow buffers are limited to a certain size (e.g. currently 8K),
and the number of bits per character cell is fixed (16bit). However, the
driver may choose how to use these bits, e.g. if it wants 9 for glyph code,
7 for colors, etc.

> I do realize that my idea of
> making the shadow buffer a unicode buffer will not work since you might
> need more than one unicode to make one characters.

Correct. All you can do is to store font indices, which is what KGI does.
Of course the back-translation (e.g. for selections) into UNICODE is not
one-to-one, but there isn't much you can do about.

> So it has a be some
> type of glyph index. Do you use a seperate attribute buffer then?

No. All info (attributes, glyph position, etc) is stored in a kgi_u16_t array.
If there ever arises a urgent need for 32bit (e.g. 24bit glyph code, 8 bit
colors)
KGI can be easily extended to support 32bit wide buffers.

However, I don't see this need, as with about 500 Glyphs you can cover most
of the european languages (german, greek, ...) sufficiently. If you need
console support for Eastern (e.g. except Korean) languages, you are better of
to not implement this inside the kernel. Mainly because input requires a
dictionary search sometimes.

			Steffen

_______________________________________________________________________________
Steffen Seeger                              mailto:se...@ph...
TU-Chemnitz                                      http://www.tu-chemnitz.de/~sse

Re: Console

[James S. <jsi...@ac...>]

> My name is Kirti Desai. I am trying to write a new
> console driver.

No problem. What are you looking to do?

> * Is console related to a tty_driver? 

A tty is a posix terminals which cover alot of different types of
hardware. Everything from modems to serial consoles to video terminals
which use hardware like VGA video cards. If you are talking about 
video consoles (VTs) then they are a subset of ttys types. Remember their
are many types of ttys.

> We have two functions register_console and register_tty_driver? How are
> these two related? 

As for register_console you are talking about the system console.  
This is the console where things like kernel errors printed via printk
go. Note any kind of tty can be a system console. 

> In tty_init(tty_io.c) we register dev_tty_driver  and
> dev_syscons_driver?
> In console_init(tty_io.c) con_init(tty_io.c for
> CONFIG_VT=y) is 
> called. It registers console_driver and calls
> register_console(vt_console_driver).

This is registering a video terminal (VT) as a system console. Thus
printks will be sent to this terminal.

> * To do printk we need not open the /dev/console file?

Printk is handled inside the kernel. Take a look at linux/kernel/printk.c.
What do you mean by open? Via userland ?

> What I found was
> start_kernel(init/main.c) calls console_init and
> kernel_thread(init..)
> which calls init is called later which opens
> /dev/console.

Right. First we initalize the the basic termios structure and startup up 
hardware that can be started early. Later on other types of hardware which
depend on other subsystems running such as the PCI bus are called in
tty_init. Next /dev/console is opened. I believe this is the first opened
file in a linux system thus its file descriptor is 0. It is stdin. With
dup we set stdin, stdout, and stderr all equal to each other. 

 if (open("/dev/console", O_RDWR, 0) < 0)
     printk("Warning: unable to open an initial console.\n");
 (void) dup(0);
 (void) dup(0);

Then init is started after the kernel is done. It is the father of all
process. It uses the file /etc/inittab to setup all your consoles. 

> * How are stdin/stdout/stderr mapped to the tty's?

See above.

> What  want to know
> if that when we call say printf in user
> program,finally write(1,..) gets called which calls
> tty_write ,con_write. How does tty come in
> picture?

printf sends by default data to stdout whcih goes to /dev/console threw
the tty layer to con_write if it is VT.

> * What are /dev/console /dev/tty0, /dev/tty1 files?

/dev/console is the system console (where the printk go). /dev/tty
represents the current console for the current process. If I'm on a serial
console and do a echo "hi" > /dev/tty it will print out on the screen. If
I'm using a console run by fbcon it will do the same thing. Now /dev/tty0
represents the current foreground video console. If you do a echo to
/dev/tty0 it will be printed on the current video display. /dev/tty1 and
so on represent the virtaul consoles for VTs. Those are the consoles you
switch to when you press Alt-F[1-6].   

> Any pointer to documentation ..would be  of great
> help?

Not alot. This is a linux console project which just started. See the
address in my sig.

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Re: [linux-fbdev] Console on frame-buffer... how?

[James S. <jsi...@ac...>]

> > I assume you pass `console=ttyS0' (or something like that) to the kernel?
> In
> > that case no console messages are sent to the console on the frame buffer
> > device.
> 
> I dont pass anything to the kernel... in the PPC distribution, a kernel
> parameter is CONFIG_SERIAL_CONSOLE that initializes a console using the
> serial driver. No parameters are required. It only add a call in
> 'console_init()'.
> 
> I looked at the sources, but i still think there's something wrong on
> console initialization (this is 'main.c'):
> 
> CONSOLE_INIT
>     CON_INIT (vt initialization)
>     CONSOLE_8xx_INIT (serial console on the PPC)
> TTY_INIT
>     RS_8xx_INIT (serial driver on the PPC)
> VIDEO_SETUP

The above is right. console_init does early terminal setup and initializes
hardware that can be called that early. con_init and console_8xx_init are
examples of drivers that can be called early. Later tty_init initialized 
hardware that needs things like pci to be setup for it. That is why fbcon
is called later. Because fbcon can not be called early a dummy console is
setup first. Then fbcon takes over this dummy console.

> Then, i see that VT is not registered because at CON_INIT, the functions
> tests the value of 'conswitchp' before running any 'con_startup' function,
> and it is NULL. Who should initialize it if this function has to be called
> before any other one that initializes the framebuffer? I cannot change
> virtual terminal because VT is not registered, then there is something wrong
> on this, isnt it?

No. conswitchp is setup in some c file in the arch directory. For the PPC
look at linux/arch/ppc/kernel

chrp_setup.c:   conswitchp = &dummy_con;
pmac_setup.c:   conswitchp = &dummy_con;
prep_setup.c:   conswitchp = &vga_con; 

> > You may want to try `console=tty0 console=ttyS0' to get console output on
> both
> > the frame buffer device and the serial console. Input for the console will
> > come from the last console= argument (i.e. the serial console, cfr.
> > Documentation/serial-console.txt).
> 
> Sorry, but it still doesnt work!

Did you enable Support for console on virtual terminal after the Virtual
terminal option ?

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Re: how to map the console to framebuffer on boot?

[James S. <jsi...@ac...>]

> i'm working on an embedded PowerPC processor with a hardhat linux kernel
> (2.2.14). This linuxbox only has a serial_console because it has no
> keyboard.
>
> now, i wrote a framebuffer device that seems to work, but i dont know
> how to activate on it a console on the startup. This is what happens on
> the kernel startup (i also use the serial console):
>
> 'fb_init' initializes my fb driver, opens 63 VC and calls
> 'take_over_console', but it does nothing because 'vc_cons[i].d' is NULL.
> Then, it displays 'Console: switching to frame buffer device', but it is
> not true!
>
> Then, i tried some experiments, and i got:
>
> echo "pippo" > /dev/tty  ==> /dev/tty: Device not configured (in
> /proc/tty/drivers i found "5 0 system:/dev/tty")

This is really wrong. /dev/tty should be pointing to the system console.
Something should come out on the serial console.

> echo "pippo" > /dev/tty0  ==> /dev/tty0: No such device (in
> /proc/tty/drivers i found "4 0 system:vtmaster")
>
> Now, what i wrong?

No keyboard. Also have you tried on the lilo prompt console=ttyS1 console=tty0
This will tell the system to use both the video console (fbcon) and the
serial console. Without a keyboard no input can go threw fbcon.

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Re: New VGA console coming

[James S. <jsi...@ac...>]

Sorry I have been away but I was sick. 

> The console layer handles the glyph information to be displayed using full
> UNICODE-2.0 (or it's successor, the ISO norm for 32bit character codes).

Same here. By the way I have expanded the conmakehash.c program to
generate a unicode->font map of up to 65,536 characters. 

> Using the font and attribute information, this is translated into a
> font-specific code, which is stored in the shadow buffer using a pattern
> defined by the graphics driver only. Then the graphics driver does the
> rendering from the shadow buffer, it never gets involved with the
> glyph-code -> font position translation. This way it is up to the graphics
> driver, which attributes and how many font characters it supports.

So the shadow buffer is driver specific. Couldn't the shadow buffer get
really huge for some types of hardware ? I do realize that my idea of
making the shadow buffer a unicode buffer will not work since you might
need more than one unicode to make one characters. So it has a be some
type of glyph index. Do you use a seperate attribute buffer then?

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Re: [linux-fbdev] Console on frame-buffer... how?

[Geert U. <ge...@li...>]

On Tue, 12 Sep 2000, Paolo Scaffardi wrote:
> Then, i see that VT is not registered because at CON_INIT, the functions
> tests the value of 'conswitchp' before running any 'con_startup' function,
> and it is NULL. Who should initialize it if this function has to be called
> before any other one that initializes the framebuffer? I cannot change
> virtual terminal because VT is not registered, then there is something wrong
> on this, isnt it?

Make sure to setup the dummy console early in your startup code (cfr.
{chrp,pmac}_setup.c):

    conswitchp = &dummy_con;

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@li...

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

Re: [linux-fbdev] Console on frame-buffer... how?

[Paolo S. <ar...@ti...>]

Hi Geert!

> I assume you pass `console=ttyS0' (or something like that) to the kernel?
In
> that case no console messages are sent to the console on the frame buffer
> device.

I dont pass anything to the kernel... in the PPC distribution, a kernel
parameter is CONFIG_SERIAL_CONSOLE that initializes a console using the
serial driver. No parameters are required. It only add a call in
'console_init()'.

I looked at the sources, but i still think there's something wrong on
console initialization (this is 'main.c'):

CONSOLE_INIT
    CON_INIT (vt initialization)
    CONSOLE_8xx_INIT (serial console on the PPC)
TTY_INIT
    RS_8xx_INIT (serial driver on the PPC)
VIDEO_SETUP

Then, i see that VT is not registered because at CON_INIT, the functions
tests the value of 'conswitchp' before running any 'con_startup' function,
and it is NULL. Who should initialize it if this function has to be called
before any other one that initializes the framebuffer? I cannot change
virtual terminal because VT is not registered, then there is something wrong
on this, isnt it?

> You may want to try `console=tty0 console=ttyS0' to get console output on
both
> the frame buffer device and the serial console. Input for the console will
> come from the last console= argument (i.e. the serial console, cfr.
> Documentation/serial-console.txt).

Sorry, but it still doesnt work!

Please reply me because i have no idea on what to do now!!!

=-> MAY THE PINGUIN B WITH U! <-=

Best regards,
Paolo Scaffardi (ar...@ti... - psc...@su...)
AIRVENT SAM s.p.a. - RIMINI (ITALY)
Tel. 0541 383294 - Fax 0541 387086

how to map the console to framebuffer on boot?

[Paolo S. <ar...@ti...>]

hi there,
i'm a newbie here, and i have lots of unanswered questions that i'd like to
make...

i'm working on an embedded PowerPC processor with a hardhat linux kernel
(2.2.14). This linuxbox only has a serial_console because it has no
keyboard.

now, i wrote a framebuffer device that seems to work, but i dont know how to
activate on it a console on the startup. This is what happens on the kernel
startup (i also use the serial console):

'fb_init' initializes my fb driver, opens 63 VC and calls
'take_over_console', but it does nothing because 'vc_cons[i].d' is NULL.
Then, it displays 'Console: switching to frame buffer device', but it is not
true!

Then, i tried some experiments, and i got:

echo "pippo" > /dev/tty  ==> /dev/tty: Device not configured (in
/proc/tty/drivers i found "5 0 system:/dev/tty")

echo "pippo" > /dev/tty0  ==> /dev/tty0: No such device (in
/proc/tty/drivers i found "4 0 system:vtmaster")

Now, what i wrong?
Is that recold field value right, or someone should set it before the
fb_init?
How a fb can become a console when booting?

it's not clear to me... i looked everywhere but that code it's very...
confusing!

Thanx again...

Best regards,
Paolo Scaffardi (ar...@ti... - psc...@su...)
AIRVENT SAM s.p.a. - RIMINI (ITALY)
Tel. 0541 383294 - Fax 0541 387086

how to map the console to framebuffer on boot?

[Paolo S. <ar...@ti...>]

hi there,
i'm a newbie here, and i have lots of unanswered questions that i'd like to
make...

i'm working on an embedded PowerPC processor with a hardhat linux kernel
(2.2.14). This linuxbox only has a serial_console because it has no
keyboard.

now, i wrote a framebuffer device that seems to work, but i dont know how to
activate on it a console on the startup. This is what happens on the kernel
startup (i also use the serial console):

'fb_init' initializes my fb driver, opens 63 VC and calls
'take_over_console', but it does nothing because 'vc_cons[i].d' is NULL.
Then, it displays 'Console: switching to frame buffer device', but it is not
true!

Then, i tried some experiments, and i got:

echo "pippo" > /dev/tty  ==> /dev/tty: Device not configured (in
/proc/tty/drivers i found "5 0 system:/dev/tty")

echo "pippo" > /dev/tty0  ==> /dev/tty0: No such device (in
/proc/tty/drivers i found "4 0 system:vtmaster")

Now, what i wrong?
Is that recold field value right, or someone should set it before the
fb_init?
How a fb can become a console when booting?

it's not clear to me... i looked everywhere but that code it's very...
confusing!

Thanx again...

Best regards,
Paolo Scaffardi (ar...@ti... - psc...@su...)
AIRVENT SAM s.p.a. - RIMINI (ITALY)
Tel. 0541 383294 - Fax 0541 387086

Re: New VGA console coming

[Steffen S. <se...@ph...>]

James Simmons wrote:

>    Now for the next step in the evolution of the console system. Well I
> made a discovery today. Their exist vga cards (JEGA, AX-VGA) that can
> display asian characters. I also discovered some old BIOS calls that might
> lead to figuring out if double byte fonts could be supported. The BIOS
> call int 10h, ah=18h, al=0,1 which can get/set a font pattern. This leads
> to me to believe it might be possible for VGA to support double byte
> fonts. This of course would mean a rethinking of how the shadow buffer
> should be done. Any ideas anyone?

Designing the KGI console system, I had a similar problem. The solution
I have implemented is not to make any assumption about how the graphics
card stores the displayed text. Basically it works this way:

The console layer handles the glyph information to be displayed using full
UNICODE-2.0 (or it's successor, the ISO norm for 32bit character codes).
Using the font and attribute information, this is translated into a
font-specific code, which is stored in the shadow buffer using a pattern
defined by the graphics driver only. Then the graphics driver does the
rendering from the shadow buffer, it never gets involved with the 
glyph-code -> font position translation. This way it is up to the graphics
driver, which attributes and how many font characters it supports. 

However, due to the current (VGA compatible) shadow buffer layout used by
KGI only 256 font characters are possible.

Hope this is of help,

			Steffen

_______________________________________________________________________________
Steffen Seeger                              mailto:se...@ph...
TU-Chemnitz                                      http://www.tu-chemnitz.de/~sse

Email address changes

[James S. <jsi...@ac...>]

Hi folks.

   To let you know my email address is going to change. My UB account will
be deactivates in 14 days. This means I can't even dail in. My email will
be queued up at linux-fbdev.org and when I get to work Oct 2 I will read
it all :-( So you will not here a response from me the last week of
Sept. Also I will be busy moving which is why I have been quiet recently.
My new email address will most likely be jsi...@su.... Please DONT
email me at that address. Use my permanent jsi...@li... email
account. Thank you.

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

News [Unicode support]

[James S. <jsi...@ac...>]

Hi!
  
   I see lots of posting last month :-) Shows the project is starting to
get a little attention. We still are at the alpha level of code and have
much to do yet. By alpha I mean we don't have all the sub systems running
(like fbdev). Here is what needs to be done yet:

1) Fix that reboot crashing. When you press alt-ctrl-del it reboots fine
   except if you press the keyboard durning reboot. I added code to prevent 
   more event packets from coming in. This stopped the instant crashed 
   after pressing ctrl-alt-del but it still crashes if you press any key
   while rebooting. Still working on it.

2) I see alot of global variables in keyboard.c. Have to see if it is 
   safe to use them. 

2) Better unicode support. Thanks to a patch from Edmund I'm working on a 
   and testing unicode code for selections. Now you will be able to cut
   and paste w_chars (when in VC_UNICODE mode) as well as regular 8 bit
   chars.

2) Documentation on how a key press is turned into a font displayed on
   the screen. I have been following threw that code recently. Boy was
   that fun. 
 
3) New screenbuf layout. I have been talking to Paul Williams about
   unicode and how to handle it. Right now the console system has a
   VC_UNICODE flag to tell the console system to use UTF-8 and peice 
   together bytes to display w_chars. The problem is in this mode you
   lose the C1 control characters (Stuff like echo -e '\033[?17;0;64c').
   He pointed out to me how Unicode 3.0 hints on how to handle this. That
   is to extend the 8 bit squences such as:
 
   ESC  [   ?   5   h
   1b  5b  3f  35  68

   out to:

   ESC   [    ?    5    h
   001b 005b 003f 0035 0068

   I might arrange the screenbuf to be a UCS-2 buffer. Right the console
   system trnaslates the unicode values into the native VGA text hardware
   layout in the screenbuffer. I might remove that step and have vgacon
   translate it itself. 

4) Work out the keymap per VT or possibly VC thing. 

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Re: keyboard

[James S. <jsi...@ac...>]

> > Well that is a little more complex. I seen another posting about this.
> 
> Anyone got any more info on this?

XFree86 X servers are a little different than other peices of software.
They talk directly to the hardware bypassing the kernels resource
management system. XFree86 3.3.6 has no really multihead management.
XF4.0 sort of does but it does it on its own instead of using the OS to
help it out. 

> My BIOS has AGP first (the matrox is on PCI), and it works quite
> happily (+xinerama, fb tools...) - only error message  is:
> "Video BIOS not detected in PCI space!"
> "Attempting to read Video BIOS from legacy ISA space!"
> But I guess that's something for my BIOS setup. (shadow rom or something)

Strange. I don't think this has anything to do with linux.

> It is not crucial, but it would be nice (19" instead of 14"), in the meantime
> I'll boot in straight text mode, but then I can't use X (don't ask me why my
> video card needs a framebuffer to be initialised properly for X...)

That is strange. XFree86 usually works better without fbdev. Is this what
your talking about?

> I thought vesa or matrox did (or maybe it was an older version, or maybe
> I am just going insane again)
> The USB is broken too.

Must be a older CVS download. Right now the fbdev is broken because of the
massive rewrite it needs.

> So how can I switch/use to the second head?
> Or do I just attach a keyboard to the second head?
> - no quite clear in *my* head yet, but getting there.

For a fully function VT you need one display and one keyboard. Each VT
acts independent of each other. This is done for security reasons. Imagine
at work on have one multihead linux server. Now from the server each
keyboard and monitor with extensions could run into different offices.
Here you don't want to VT switch to a terminal belonging to someone else
in some other office. 

> > > > Look at ruby CVS code for xfree86.
> > > Where is this tree?
> > Grab the ruby tree and right at the top you see a xfree86 directory.
> 
> I still don't see it! (ls -lR | grep -i xfree => nothing)
> [antoine@shami ruby]$ ls
> AGAINST-2.4.0-test7  CVS  app-patches  linux  utils

Remove the current ruby tree you have to download a new tree. I have 

AGAINST-2.4.0-test7  CVS  app-patches  linux  utils  xfree86   

> docs what docs. ;-(
> Do you think there is any chance of getting the Xfree86 people to point me in
> the right direction? Or spare me a few minutes (like you guys did).

Ask them. They have several mailing list. 

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net

Re: CVS updates/VGA mode setting

[Petr V. <VAN...@vc...>]

On  1 Sep 00 at 9:22, James Simmons wrote:
> > > What I mean is read the values from the vga hardware instead of from
> > > screeninfo which is provided by the firmware. 
> > 
> > If you want to support only standard modes (text, cfb8, vgaplanes, 
> > vga8planes), it is easy. But there is couple of more or less nonstandard
> > modes too. 2 (1plane) or 4-color (2planes) modes, alphanumeric mode with
> > character generator disabled (gives you `spectrum like' attribute + 8pixels).
> > What to do with them? Let them fall through and report something similar
> > for them?
> 
> Wow. I never even seen any spectrum modes in any books. Where did you
> learn about it?

It must work that way... I was asking myself why there are two alphanumeric-
-graphics switches. One says whether pass data through character generator,
second says whether it is data+attr, or 4plane data...

Another solution is using documented interface and simple create 1 line
tall font with 1:1 mapping (0->0, 1->1,... 255->255).  
> > If you are talking only about recognition of currently used videomode
> > at startup, it should be possible, and possibly trivial. I have
> > (in mgadump) code which adds `correct' constants and sums correct
> > bits together to get all picture sizes... Should I port that code
> > to Linux and send it to you?
> 
> Pretty much I'm talking about recongnition of currently used videomode but
> it would also be nice to support other "non-standard modes". I would like
> to see your code :-)

OK, I'll port it during weekend.
                                                    Petr
                                                    

Re: keyboard

[Petr V. <VAN...@vc...>]

On  1 Sep 00 at 9:15, James Simmons wrote:
> > Also, when using ruby, it boots off the g200 (usually boots off the r128)
> > and the kernel boot messages in dmesg do not contain whatever was
> > on the console before it switched to the g200, where can I get these early
> > messages? And why did they disappear!?
> 
> Right now with ruby only vgacon, mdacon, and nvida text mode are
> supported. The fbdev layer is incomplete. I have found matrox cards
> usually like to be booted first. I know of no way to make one card boot
> before another. Petr, you have any ideas ?

I'm using my 'pciorder' patch I sent around couple of times... 
'pciorder=1:00.0,0:0b.0,0:0f.0' and it's done. And with
old fbmem you can set order with 'video=fbdev_name:' option...
matroxfb itself does not allow you to reorder different MGA devices in
box - they are initialized in PCI order (which can be changed with
my pciorder patch which others do not like...).
                                     Petr
                                    

Re: keyboard

[Antoine M. <an...@go...>]

> I need to VT switch to change which X is active, I guess
> > tha's because they share the same keyboard. If I get one of the Xs to use
> > the usb keyboard, will they both be running concurrently then? Or do I need
> > something extra...
>
> Well that is a little more complex. I seen another posting about this.

Anyone got any more info on this?

> Right now with ruby only vgacon, mdacon, and nvida text mode are
> supported. The fbdev layer is incomplete. I have found matrox cards
> usually like to be booted first.

My BIOS has AGP first (the matrox is on PCI), and it works quite
happily (+xinerama, fb tools...) - only error message  is:
"Video BIOS not detected in PCI space!"
"Attempting to read Video BIOS from legacy ISA space!"
But I guess that's something for my BIOS setup. (shadow rom or something)

> I know of no way to make one card boot
> before another. Petr, you have any ideas ?

It is not crucial, but it would be nice (19" instead of 14"), in the meantime
I'll boot in straight text mode, but then I can't use X (don't ask me why my
video card needs a framebuffer to be initialised properly for X...)

>   How do you get the fbcon layer to even compile for ruby? I didn't think
> it compiled.

I thought vesa or matrox did (or maybe it was an older version, or maybe
I am just going insane again)
The USB is broken too.

> Also alot of the fbcon layer has global variables meaning it
> not really multihead safe. I'm rewriting that code to make it really
> multihead safe but I'm not finished.
>    The reason con2fb doesn't work is because with the new console
> system it automatically allocates 16 VCs to each head. You don't need to
> do con2fb because it should be done automatically for you. With your setup
> with the fbcon layer you should have 2 consoles come up for you automatically.
> The console code also prevents things like VC switching to a different
> head than your on.

So how can I switch/use to the second head?
Or do I just attach a keyboard to the second head?
- no quite clear in *my* head yet, but getting there.

> > > Look at ruby CVS code for xfree86.
> > Where is this tree?
> Grab the ruby tree and right at the top you see a xfree86 directory.

I still don't see it! (ls -lR | grep -i xfree => nothing)
[antoine@shami ruby]$ ls
AGAINST-2.4.0-test7  CVS  app-patches  linux  utils

> Go for it. I noticed the lack of docs myself. I atempted to read their
> code but it was such a mess I gave up.

docs what docs. ;-(
Do you think there is any chance of getting the Xfree86 people to point me in
the right direction? Or spare me a few minutes (like you guys did).

Many Thanks
Antoine

Re: CVS updates/VGA mode setting

[James S. <jsi...@ac...>]

> > What I mean is read the values from the vga hardware instead of from
> > screeninfo which is provided by the firmware. 
> 
> If you want to support only standard modes (text, cfb8, vgaplanes, 
> vga8planes), it is easy. But there is couple of more or less nonstandard
> modes too. 2 (1plane) or 4-color (2planes) modes, alphanumeric mode with
> character generator disabled (gives you `spectrum like' attribute + 8pixels).
> What to do with them? Let them fall through and report something similar
> for them?

Wow. I never even seen any spectrum modes in any books. Where did you
learn about it?
 
> If you are talking only about recognition of currently used videomode
> at startup, it should be possible, and possibly trivial. I have
> (in mgadump) code which adds `correct' constants and sums correct
> bits together to get all picture sizes... Should I port that code
> to Linux and send it to you?

Pretty much I'm talking about recongnition of currently used videomode but
it would also be nice to support other "non-standard modes". I would like
to see your code :-)

Re: keyboard

[James S. <jsi...@ac...>]

> > This is for a single PS/2 keyboard. What do you see for your setup?
> 
> similar stuff. I'll send the logs later if needed.
> but from what you said below, I understand it means that my primary
> keyboard (ps2) is bound to the console and won't be in /dev/input/event,
> is this correct?

No. For every input device a /dev/input/event should appear if you
are using devfs. Otherwise you have to create the device file yourself
(man mknod). The console system behave indpendent of the userland input
interface. This is a more powerful approach. 

> Actually, I have 2 displays (rage128 agp and g200 pci), 2 keyboards (ps2
> and usb) and 2 rodents (ps2 and usb).

Cool. 

> And I wouldn't mind running concurrently 2 full X desktops off 1 box...
> (could even run q3 multiplayer off that box!?)
> When I start 2 X servers (both using the ps2 keyboard since I couldn't
> work it out) I need to VT switch to change which X is active, I guess
> tha's because they share the same keyboard. If I get one of the Xs to use
> the usb keyboard, will they both be running concurrently then? Or do I need
> something extra...

Well that is a little more complex. I seen another posting about this.

> Also, when using ruby, it boots off the g200 (usually boots off the r128)
> and the kernel boot messages in dmesg do not contain whatever was
> on the console before it switched to the g200, where can I get these early
> messages? And why did they disappear!?

Right now with ruby only vgacon, mdacon, and nvida text mode are
supported. The fbdev layer is incomplete. I have found matrox cards
usually like to be booted first. I know of no way to make one card boot
before another. Petr, you have any ideas ?

> Then, if I try to con2fb to the r128, it does not come on (but it works on
> test7).
> Framebuffer tools (fbset, fbi or fbtv...) all go to the g200, even when I
> specify
> -c /dev/fb1 (the r128), now that's scary! 
> Also, on test7 I need to con2fb to the r128 before I can use it as a
> framebuffer
> device (all modules already loaded...), otherwise the fb mode that gets set is

  How do you get the fbcon layer to even compile for ruby? I didn't think
it compiled. Also alot of the fbcon layer has global variables meaning it
not really multihead safe. I'm rewriting that code to make it really
multihead safe but I'm not finished. 
   The reason con2fb doesn't work is because with the new console
system it automatically allocates 16 VCs to each head. You don't need to
do con2fb because it should be done automatically for you. With your setup
with the fbcon layer you should have 2 consoles come up for you automatically.
The console code also prevents things like VC switching to a different
head than your on.

> > Look at ruby CVS code for xfree86.
> 
> Where is this tree?

Grab the ruby tree and right at the top you see a xfree86 directory.

> > looked it didn't support USB keyboards or /dev/input. Which X server are
> > you running.
> 
> 4.0.1 from CVS (but I must have an old 3.3.6 tree somewhere)
> > XF4.0. With XFree86 it usually takes them 6 months to support something.
> 
> what about doing it myself?
> I was looking around for docs on xinput and the whole input stuff in X4...
> and I didn't find much info.

Go for it. I noticed the lack of docs myself. I atempted to read their
code but it was such a mess I gave up.

MS: (n) 1. A debilitating and surprisingly widespread affliction that
renders the sufferer barely able to perform the simplest task. 2. A disease.

James Simmons  [jsi...@li...]               ____/| 
fbdev/console/gfx developer                             \ o.O| 
http://www.linux-fbdev.org                               =(_)= 
http://linuxgfx.sourceforge.net                            U
http://linuxconsole.sourceforge.net