linuxconsole-dev

Dual Monitor graphics boards for multi-user local X

[Yitzhak B. G. <yit...@ya...>]

IMHO the most effective way to build ruby multi-user local X Windows and =
maximize concurrent users is by using dual=E2=80=91port graphics boards.

Aivils had indicated that dual monitor boards like our nVidia based =
boards are not suited for ruby because the chip has only one PCI bus =
address. That means one cannot define separate ruby X Servers using =
=E2=80=9Cscreen 0=E2=80=9D and =E2=80=9Cscreen 1=E2=80=9D.

If anyone has a solution to this issue, please inform us.

We need a dual chip graphics board, preferably nVidia based, or another =
Linux supported chip. So far, the only dual GPU boards I know of are =
very expensive. Where can we find inexpensive dual GPU chip PCI boards?

Thanks,

Yiztahk Bar Geva

=20

Re: Announce: brief HowTo in german (GAMES)

[Aivils <ai...@un...>]

On Wednesday 18 August 2004 22:23, Svetoslav Slavtchev wrote:
> I also try to give my debian
> > pages to Andreas Schuldei as soon as the low latancy patch is tested.

Also 2.6 note for gamers.
Low latency will not help for multiple game instances.
Normal 2.6 tack scheduler will give very long timeslices for each
process, aprox 200ms. As result if You start two copies of games,
then both games became unplayable choppy, tuxracer included.

Solution is
nice +19 game-bin
This commandline guaranted 10ms rescheduling for game-bin,
so in my end 3 simultaneous games run smooth.
2.6 task scheduler is customized for desktop apps, which is bad
for simultaneous 100% CPU tasks. According code author Con
Kolivas it must be so and cannot be changeable.

Aivils

Re: Announce: brief HowTo in german

[Svetoslav S. <sv...@gm...>]

> On Sunday 15 August 2004 13:08, Svetoslav Slavtchev wrote:
> > > Moin,
> >
> > Hi Michael,
> 
> Moin,
> 
> thanks for your answer.

not a problem :-)

> > > http://hyaden.dyndns.org/HOWTO/lmu
> >
> > 3 notes at first look
> > (well one is not really related to the contents)
> >
> > 1.)
> > i think the part about the nvidia driver
> > is totally irrelevant, and maybe even a bit confusing
> > do you really need to add a script for recompiling it ?
> 
> The HowTo is very specific to that what we did until now. With other
> words: At 
> the moment it is not more than a 'what we did until now'. We will become
> more 
> generic later. I try test more graphic cards. I also try to give my debian
> pages to Andreas Schuldei as soon as the low latancy patch is tested.

ok, 
but may be still ....
using low latency is not (deeply) related/required by (b)ruby
you might look for ll hints/links at linux-sound ml

 
> > 2.)
> > i really think it's a lot better to use display manager
> > (gdm to be presize)
> > instead of adding additional users and using a
> > bash script to start all the Xservers for each user
> 
> That depend. I agree to you if the PC should be a normal desktop for
> several 
> users. But I want to use it at a conference like a kind of 'internet
> cafe'. 
> The conference participants want to read news and mails, so an 'autostart'
> X 
> is the best for them. Maybe we alter this to KDEs kiosk mode.
> Anyhow: in later versions of the howto several methods will be described.

guess who sugested the note about i-net cafe's and school pools
at bruby's page :-)

IIRC both gdm and kdm support autologin,
and IMHO it's a lot cleaner to use display manager
instead of shell script to bring up/restart X

 
> > 3.) (the not related question)
> >
> > why is no one giving me any feedback for the howto,
> > and starts it's own one, IMHO it's confusing to have
> > a lot of howto's on the same topic, with some small
> > differencies
> 
> Well, this is an important question. Your HowTo was very helpful to me!
> But 
> many people here don't speak english! They need a german text.
> Maybe we should translate your document!

good luck :-)
just kidding, but i don't have the time
and i as i'm not a native german speaker
it'll be pretty time consuming for me

i'll be really happy if someone translate it
in other languages thought


> > the one thing i know, is that my howto is simply huge
> > but i don't see how to make it smaller without
> > leaving some functionality
> 
> That's the point! But it is all-embracing, and therefore useful for
> everybody.
> 

yes, but also scary for newbies :(
i really don't know what to do about this

on a mandrake PC it's almost as easy as
install kernel and modified X using a package manager
adjust configuration 
- for a 2 user system this could be limited just to 
  editing gdm : add aditional X server
  and Xf86Config-4/xorg.conf (add two new layouts)
- for more users hotplug configuration is also needed 
  and it's getting more complicated
(for debian with Andreas packages, the same probably applies too)

best,

svetljo

-- 
NEU: WLAN-Router für 0,- EUR* - auch für DSL-Wechsler!
GMX DSL = supergünstig & kabellos http://www.gmx.net/de/go/dsl

Re: Re: Announce: brief HowTo in german

[Svetoslav S. <sv...@gm...>]

> > 2.) 
> > i really think it's a lot better to use display manager (gdm to be
> > presize) instead of adding additional users and using a bash script to
> > start all the Xservers for each user
> 
> As for me any display manager is suitable. I use time to time ruby-2.6
> and Miguel Freitas solutions with "xdm" display manager, so i use
> IsolateDevice and evdev patches both. In both case i use "-delay" options
> (come from evdev patch from Miguel Freitas) to specify a right order of
> Xservers loading (so gdm is not required). I think its wise to add
> "-delay" option to IsolateDevice patch to.

hm, if you were using gdm you wouldn't need -delay
as gdm sort of use "-delay"
e.g.
start X0 , sleep 1s, start X1, sleep 1s, start X2,sleep 1s .....
or .., stop X2, sleep 1s, stop X1, sleep 1s, stop X0
have in mind that i'm not sure about the exact delay

IMHO a lot more important is to get "-vtshift [n] " option,
but as i'm not a X hacker ... :(


> > 3.) (the not related question)
> > why is no one giving me any feedback for the howto, and starts it's own 
> > one, IMHO it's confusing to have a lot of howto's on the same topic,
> > with some small differencies
> 
> 
> i can send you my kernel/XOrg config with some comments 
> 

please do so 
:-)

i haven't got a single comment for at least 6months
( well ever since the addition the notes about gentoo)

best,

svetljo

-- 
NEU: WLAN-Router für 0,- EUR* - auch für DSL-Wechsler!
GMX DSL = supergünstig & kabellos http://www.gmx.net/de/go/dsl

Re: Help with Linux Multihead

[Aivils <ai...@un...>]

> =A0=A0=A0 Thanks, Mr. Aivils,
> =20
>  =A0=A0=A0 I tried several configurations, but no one works.
> =20
>  =A0=A0=A0 If I put Single Card option or PrefBusID option in XF86Config,=
 when I start the second xserver the first one go down; when I start the th=
ird xserver, the second go down and etc. If I don't use these options, all =
monitors show xserver, but only the last answer mouse and keyboard events. =
Others monitors stay frooze!

All right.

>  =A0=A0=A0 That last configuration I tried was a Mandrake 9.2 with 2.4.25=
 backstreet ruby kernel and binary patched kernel from your page (http://st=
artx.times.lv/XFree86-430-prefbusid3.bz2).=20
>  =A0=A0=A0=20
>  =A0=A0=A0 I used kdb.conf, input.agent, input.rc, mouse.conf , like you.=
 I have the vc's working like in howto:
> =20
>  [root@1 xc]# cat /proc/bus/console/*/*
>  isa0060/serio0/input0
>  usb-00:10.1-1/input0
>  usb-00:10.1-2/input0
>  usb-00:10.0-1/input0
>  usb-00:10.0-2/input0
>  =A0=20
>  =A0=A0=A0 And links were created in /dev/input/mouse?br.
> =20
>  =A0=A0=A0 I have 5 Ati Rage 128 PCI cards and used the comands:
> =20
>  =A0 261=A0 startx -- /usr/X11R6/bin/X4 -layout :0 X5 vt7
>  =A0 262=A0 startx -- /usr/X11R6/bin/X3 :0 -layout X4 vt20
>  =A0 263=A0 startx -- /usr/X11R6/bin/X2 :0 -layout X1 vt17
>  =A0 265=A0 startx -- /usr/X11R6/bin/X1 :0 -layout X2 vt18
>  =A0 266=A0 startx -- /usr/X11R6/bin/X0 :0 -layout X3 vt19
>  =A0=A0=A0=20
>  =A0=A0=A0 Sometimes I use a NVIDIA AGP instead of an ATI.
> =20
>  =A0=A0 Please, what can be wrong ?

Some olders XFree86 drivers will not support multiheading,
multiple adapters.
Please try hackvideo, secondary method of xf86 starting
without prefbusid
http://www.schuldei.org/aivils/eng-xf.html

>  =A0=A0 Can you help me ? =A0 =20

I never tried ATI hardware. May be any other tried ATI Rage ?

ATI mach 64 +  ATI mach 64 will not work.
At least may work two head system - Nvidia AGP + ATI,
where "nvidia"  closed source driver used. www.nvidia.com

If You cannot start it , then i add ATI rage 128 as incompatible
video adapters to prevent others use them.

Aivils

Re: Announce: brief HowTo in german

[Michael R. <mic...@gm...>]

On Sunday 15 August 2004 13:08, Svetoslav Slavtchev wrote:
> > Moin,
>
> Hi Michael,

Moin,

thanks for your answer.

[...]

> > http://hyaden.dyndns.org/HOWTO/lmu
>
> 3 notes at first look
> (well one is not really related to the contents)
>
> 1.)
> i think the part about the nvidia driver
> is totally irrelevant, and maybe even a bit confusing
> do you really need to add a script for recompiling it ?

The HowTo is very specific to that what we did until now. With other words: At 
the moment it is not more than a 'what we did until now'. We will become more 
generic later. I try test more graphic cards. I also try to give my debian 
pages to Andreas Schuldei as soon as the low latancy patch is tested.

> 2.)
> i really think it's a lot better to use display manager
> (gdm to be presize)
> instead of adding additional users and using a
> bash script to start all the Xservers for each user

That depend. I agree to you if the PC should be a normal desktop for several 
users. But I want to use it at a conference like a kind of 'internet cafe'. 
The conference participants want to read news and mails, so an 'autostart' X 
is the best for them. Maybe we alter this to KDEs kiosk mode.
Anyhow: in later versions of the howto several methods will be described.

> 3.) (the not related question)
>
> why is no one giving me any feedback for the howto,
> and starts it's own one, IMHO it's confusing to have
> a lot of howto's on the same topic, with some small
> differencies

Well, this is an important question. Your HowTo was very helpful to me! But 
many people here don't speak english! They need a german text.
Maybe we should translate your document!
 
> the one thing i know, is that my howto is simply huge
> but i don't see how to make it smaller without
> leaving some functionality

That's the point! But it is all-embracing, and therefore useful for everybody.

CU
-- 
|Michael Renner      E-mail: mic...@gm...  |
|D-72072 Tuebingen   Germany        ICQ: #112280325 |
|Germany             Don't drink as root!      ESC:wq

Re: How to use framebuffer(s) with ruby?

[Helge H. <hel...@hi...>]

dave giffin wrote:

>I've used backstreet ruby before with 2 regular
>XFree86 servers.
>  
>
This is possible with linux-2.6+ruby too.

>Now, I would like to be able to use an XDirectFB (an X
>server using the DirectFB framebuffer library)
>instance on one card while simultaneously using a
>regular XFree86.
>  
>
Why do you want to use the framebuffers?  X is usually faster using the
regular accelerated servers.  Or is that impossible for some reason?

>I want to be able to use a regular (non-framebuffer)
>XFree86 server on one graphics card and a framebuffer
>app on another graphics card.
>
>  
>
This is possible.

>If that is not possible, I would settle for using two
>independent framebuffer apps, one on each card (I'd
>just use two XDirectFBs)
>
>  
>
Also possible.

>How do I use ruby (linux-2.6) to setup multi-user
>framebuffers?
>  
>
First, install the hardware, i.e. your graphichs cards, monitors
mice, and keyboards.  Then compile a linux 2.5+ruby kernel, with
framebuffer support for the card you want to use.  You may want to
use framebuffer console support if you want to run regular logins on
the framebuffer when not using X.  If you use fbcon, you probably need
framebuffer support for your regular card too, so ruby will understand
that there is two consoles. IF you don't use fbcon, use the "dumbcon=1"
parameter on the kernel command line.  "dumbcon" create a input-only
console for the second keyboard.  A keyboard must hook up with
_some_ console, or it goes unused.

Install your regular xserver in the regular way, i.e. the way
it is usually installed with your distribution.  You may have to use
the prefbusid/isolatedevice/singlecard option in XF86config so the other 
card
doesn't get disabled.  (Read mailing list archives or search google
for details on that.)  If your xserver doesn't support this, compile 
your own
which is a lot of work.  The xserver in debian testing should be fine 
though.
If you get trouble, also look for the "hackvideo" solution which achieve the
same in a different way.

You may also want to verify that the framebuffer for the other display 
works.
Do that by catting a file into the framebuffer device, and verify that 
garbage
indeed appear on the other screen.
cat file > /dev/fb0
or
cat file > /dev/fb1
depending on which framebuffer device that card is.  (If you don't know,
look at dmesg output or just try the commands.) Any non-empty file will do.

Now set up the second xserver, by creating an alternate XF86Config-4 file.
Call it something different from XF86Config-4, because that filename is
used for your regular xserver. 

The "device" section should contain
Driver   "fbdev"
in order to use the X framebuffer driver, and
Option   "fbdev"   "/dev/fb0"
or
Option   "fbdev" "/dev/fb1"
depending on what framebuffer the second screen is on.

Framebuffer X does not need to know what card or what kind
of card it uses - it does everything through the kernel
framebuffer interface.  This is unaccelerated but usually fast enough
for 2D apps.  You can even use software 3D if you like - but it'll be way
too slow for anything but very simple effects.  (It is fine for playing
the frozen-bubble game :-)

When starting the second xserver, use the -xf86config parameter
to specify your alternate X config file.  Also specify what VT the
server should use.  With ruby, the second keyboard is usually on VT17.
(It might be higher up if you're using fbcon and runs a getty on vt17.  But
console switching on secondary framebuffers doesn't work too well yet, so
you probably don't do that.)

I use xdm to manage two xservers, one regular accelearated and one 
framebuffer.
My /etc/X11/xdm/Xservers look like this:

#regular, using /etc/X11/XF86Config-4, running on vt7

:0 local /usr/X11R6/bin/X :0 vt7  -deferglyphs 16 -nolisten tcp

#framebuffer, using /etc/X11/xf86config-4 running on vt17
:1 local /usr/X11R6/bin/X :1 vt17 -deferglyphs 16 -nolisten tcp 
-xf86config xf86config-4


If you want to start stuff from a script instead of xdm, use something 
like this (untested):
/usr/X11R6/bin/X :0 vt7  -deferglyphs 16 -nolisten tcp &
/usr/X11R6/bin/X :1 vt17 -deferglyphs 16 -nolisten tcp -xf86config 
xf86config-4  &



You might want to run a font server when using two xservers, for 
efficieny reasons.
If you don't then both xservers will run their own internal font server, 
but running two
font servers probably waste memory.  Using one xfs, and specifying
 FontPath        "unix/:7100"
in the X config files is probably better.

Helge Hafting

How to use framebuffer(s) with ruby?

[dave g. <giv...@ya...>]

I've used backstreet ruby before with 2 regular
XFree86 servers.

Now, I would like to be able to use an XDirectFB (an X
server using the DirectFB framebuffer library)
instance on one card while simultaneously using a
regular XFree86.

I want to be able to use a regular (non-framebuffer)
XFree86 server on one graphics card and a framebuffer
app on another graphics card.

If that is not possible, I would settle for using two
independent framebuffer apps, one on each card (I'd
just use two XDirectFBs)



How do I use ruby (linux-2.6) to setup multi-user
framebuffers?

(I can't seem to find instructions)

:)


		
__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - Send 10MB messages!
http://promotions.yahoo.com/new_mail 

Re: Planning The upgrade

[Aivils <ai...@un...>]

On Saturday 14 August 2004 19:08, jam...@em... wrote:
> I have lookd with interst at the many "how to pages" for setting up a dual 
> headed system,a nd it seems tha tmost of them were last updated when Kernel 2.4 
> and XFree86 were still the standard.
> Currently I am using 2.6 with the ALSA patches, and X.org as my Xserver. and 
> Would like to be able to upgrade my system while downgrading as little software 
> as possible.
> The most promising tutorial I have found at 
> 
> http://cambuca.ldhs.cetuc.puc-rio.br/multiuser/
> 
> include the following section.
> -Proper kernel 2.6 support
> -Aivils Stoss notified me that multiple keyboards are possible without any kind 
> of 2.6 kernel patching (see above). The trick is to request exclusive access to 
> the event device, so that keyboard handler would not receive any event. This 
> requires only small changes to the XFree86 patch (actually x.org fork as i 
> expect the old XFree86 to die silently with their new license...). Aivils 
> explains:
> 
> data[0]=1;
> if (ioctl(fd, EVIOCGRAB, data) == 0)
> printf("exclusive access granted");
> 
> "where fd is file descriptor of /dev/input/eventXX. I dont know wherethat is 
> documented.
> 
> If You will highly polish Your solution, then take a look into Debian testing 
> distro: Debian testing xf86 already includes 2.6.XX event device support.
> 
> Please check out this patch against xf86 4.3.0: 
> http://ftp.debian.org/debian/pool/main/x/xfree86/xfree86_4.3.0-7.diff.gz (grep 
> evdev). You can find this link via www.debian.org. Unfortunately i do not know 
> how to works Debian. I know only some linux-ruby testersuse this event 
> interface. xf86 event interface means usage of /dev/input/eventXX instead 
> standard /dev/ttyXX for keyboard and /dev/input/mouseXX for mouse.If You will 
> run multiple local Debian xf86 You should applay manualy Your own "TTY 
> dependency remove" patch and seems add EVIOCGRAB to disable 2nd, 3rd ... 
> keyboard presence on TTY and "xf86 PCI interference remove" patch.
> 
> Advantage: kernel remain unpatched. You can use two PS/2 keyboards under 
> 2.6.XX." 
> 
> With All of the talk of Ruby kernel patches, this almost seems too good to be 
> true...and I don't really understand where to place those lines in the patch 
> either(I am learning linux as quickly as oppurtunity permits).

I mistake little when i send mail to Miguel. Those lines already are included
in the Debian evdev patches.
* X server will use mouse and keyboard - for this You need Debian evdev
pathes. After applaying of pathes X server read mouse and keyboard events
from /dev/input/eventXX.
* X server must not stop any video adapter in system - for this need Debian
isolateDevice patch.
* X server must not open /dev/ttyXX - take a look into Miguel' s patches

If You are capable create this 3 things, then You can use "multiuser"
system with vanilla kernel (2.4 or 2.6).
It is not so hard. Seems Miguel are tired.

You are your own!

Most users of this mailing list use patched linux kernel, i hope :)

> I haven't begun yet on seting up my system, or even purchasing additional 
> components until I get my homework done on the software.
> It would also seem, with two monitors windows could easily be setup as a 
> xinerama system.  X.org is easily configured likewise (it is part most of the 
> installation tutorials I have read).  How simple can the switch be made 
> from "xinerama" to "dual-head" mode? 

Some end-users are not capable install fresh kernel from source.
Check yourself :)

> Has Hot switching (commendering the other  
> monitor while others are logged off, and being able to relinquish it to allow 
> others to log on) been explored yet?

?
Root can only kill/restart aditionals X servers. i do not know how to command
or run some programm on another X server.

Aivils

P.S. Also exsits nested X server Xnest. May be anyone will patch it and so
run nested X with separate, unnested keyboard/mouse ?

Re: Announce: brief HowTo in german

[Svetoslav S. <sv...@gm...>]

> Moin,
Hi Michael,

> I wrote a HowTo in german language. It covers the configuration for 4
> Nvidia 
> cards. It's not complete jet, but it may help already!
> 
> http://hyaden.dyndns.org/HOWTO/lmu

3 notes at first look 
(well one is not really related to the contents)

1.) 
i think the part about the nvidia driver
is totally irrelevant, and maybe even a bit confusing
do you really need to add a script for recompiling it ?

2.)
i really think it's a lot better to use display manager
(gdm to be presize)
instead of adding additional users and using a 
bash script to start all the Xservers for each user

3.) (the not related question)

why is no one giving me any feedback for the howto,
and starts it's own one, IMHO it's confusing to have
a lot of howto's on the same topic, with some small
differencies

the one thing i know, is that my howto is simply huge
but i don't see how to make it smaller without
leaving some functionality

i'll gladly accept coments and patches to the cvs version
(at tldp.org) 

best,

svetljo

PS.

i've sent a small update yesterday
including some notes about IsolateDevice and it's
inclusion in Mandrake's xorg packages and Debian Sid's xfree86 packages
and i droped some comments about the brokeness of 
linux-2.6, but there are still lot to go when i find some time

PPS.
i feel like also some doc's on multiple console users
are needed, but ...
i don't wont to blow the howto even more,
and i'm not sure what to do
also using udev with 2.6 should alow managing
usb (storage/video/audio/mice) devices for each desktop
a lot easier IMHO

PPPS.
if you have some patches, please wait till
till tldp.org updates the cvs version to the one 
i submitted yesterday v1.4

-- 
NEU: WLAN-Router für 0,- EUR* - auch für DSL-Wechsler!
GMX DSL = supergünstig & kabellos http://www.gmx.net/de/go/dsl

Announce: brief HowTo in german

[Michael R. <mic...@gm...>]

Moin,

I wrote a HowTo in german language. It covers the configuration for 4 Nvidia 
cards. It's not complete jet, but it may help already!

http://hyaden.dyndns.org/HOWTO/lmu

CU
-- 
|Michael Renner      E-mail: mic...@gm...  |
|D-72072 Tuebingen   Germany        ICQ: #112280325 |
|Germany             Don't drink as root!      ESC:wq

Planning The upgrade

[<jam...@em...>]

I have lookd with interst at the many "how to pages" for setting up a dual 
headed system,a nd it seems tha tmost of them were last updated when Kernel 2.4 
and XFree86 were still the standard.
Currently I am using 2.6 with the ALSA patches, and X.org as my Xserver. and 
Would like to be able to upgrade my system while downgrading as little software 
as possible.
The most promising tutorial I have found at 

http://cambuca.ldhs.cetuc.puc-rio.br/multiuser/

include the following section.
-Proper kernel 2.6 support
-Aivils Stoss notified me that multiple keyboards are possible without any kind 
of 2.6 kernel patching (see above). The trick is to request exclusive access to 
the event device, so that keyboard handler would not receive any event. This 
requires only small changes to the XFree86 patch (actually x.org fork as i 
expect the old XFree86 to die silently with their new license...). Aivils 
explains:

data[0]=1;
if (ioctl(fd, EVIOCGRAB, data) == 0)
printf("exclusive access granted");

"where fd is file descriptor of /dev/input/eventXX. I dont know wherethat is 
documented.

If You will highly polish Your solution, then take a look into Debian testing 
distro: Debian testing xf86 already includes 2.6.XX event device support.

Please check out this patch against xf86 4.3.0: 
http://ftp.debian.org/debian/pool/main/x/xfree86/xfree86_4.3.0-7.diff.gz (grep 
evdev). You can find this link via www.debian.org. Unfortunately i do not know 
how to works Debian. I know only some linux-ruby testersuse this event 
interface. xf86 event interface means usage of /dev/input/eventXX instead 
standard /dev/ttyXX for keyboard and /dev/input/mouseXX for mouse.If You will 
run multiple local Debian xf86 You should applay manualy Your own "TTY 
dependency remove" patch and seems add EVIOCGRAB to disable 2nd, 3rd ... 
keyboard presence on TTY and "xf86 PCI interference remove" patch.

Advantage: kernel remain unpatched. You can use two PS/2 keyboards under 
2.6.XX." 

With All of the talk of Ruby kernel patches, this almost seems too good to be 
true...and I don't really understand where to place those lines in the patch 
either(I am learning linux as quickly as oppurtunity permits).
I haven't begun yet on seting up my system, or even purchasing additional 
components until I get my homework done on the software.
It would also seem, with two monitors windows could easily be setup as a 
xinerama system.  X.org is easily configured likewise (it is part most of the 
installation tutorials I have read).  How simple can the switch be made 
from "xinerama" to "dual-head" mode?  Has Hot switching (commendering the other 
monitor while others are logged off, and being able to relinquish it to allow 
others to log on) been explored yet?

BTW, MY system specs,
Xeon 2.8 Ghz Intel, 300 GBhd, 256MB ram, intergrated intel Mother/Video/Sound 
Board, Lots of Open PCI slots, dual booting XP and ArchLinux.

Planning The upgrade

[<jam...@em...>]

I have lookd with interst at the many "how to pages" for setting up a dual 
headed system,a nd it seems tha tmost of them were last updated when Kernel 2.4 
and XFree86 were still the standard.
Currently I am using 2.6 with the ALSA patches, and X.org as my Xserver. and 
Would like to be able to upgrade my system while downgrading as little software 
as possible.
The most promising tutorial I have found at 

http://cambuca.ldhs.cetuc.puc-rio.br/multiuser/

include the following section.
-Proper kernel 2.6 support
-Aivils Stoss notified me that multiple keyboards are possible without any kind 
of 2.6 kernel patching (see above). The trick is to request exclusive access to 
the event device, so that keyboard handler would not receive any event. This 
requires only small changes to the XFree86 patch (actually x.org fork as i 
expect the old XFree86 to die silently with their new license...). Aivils 
explains:

data[0]=1;
if (ioctl(fd, EVIOCGRAB, data) == 0)
printf("exclusive access granted");

"where fd is file descriptor of /dev/input/eventXX. I dont know wherethat is 
documented.

If You will highly polish Your solution, then take a look into Debian testing 
distro: Debian testing xf86 already includes 2.6.XX event device support.

Please check out this patch against xf86 4.3.0: 
http://ftp.debian.org/debian/pool/main/x/xfree86/xfree86_4.3.0-7.diff.gz (grep 
evdev). You can find this link via www.debian.org. Unfortunately i do not know 
how to works Debian. I know only some linux-ruby testersuse this event 
interface. xf86 event interface means usage of /dev/input/eventXX instead 
standard /dev/ttyXX for keyboard and /dev/input/mouseXX for mouse.If You will 
run multiple local Debian xf86 You should applay manualy Your own "TTY 
dependency remove" patch and seems add EVIOCGRAB to disable 2nd, 3rd ... 
keyboard presence on TTY and "xf86 PCI interference remove" patch.

Advantage: kernel remain unpatched. You can use two PS/2 keyboards under 
2.6.XX." 

With All of the talk of Ruby kernel patches, this almost seems too good to be 
true...and I don't really understand where to place those lines in the patch 
either(I am learning linux as quickly as oppurtunity permits).
I haven't begun yet on seting up my system, or even purchasing additional 
components until I get my homework done on the software.
It would also seem, with two monitors windows could easily be setup as a 
xinerama system.  X.org is easily configured likewise (it is part most of the 
installation tutorials I have read).  How simple can the switch be made 
from "xinerama" to "dual-head" mode?  Has Hot switching (commendering the other 
monitor while others are logged off, and being able to relinquish it to allow 
others to log on) been explored yet?

BTW, MY system specs,
Xeon 2.8 Ghz Intel, 300 GBhd, 256MB ram, intergrated intel Mother/Video/Sound 
Board, Lots of Open PCI slots, dual booting XP and ArchLinux.

Re: User questions welcome?

[Svetoslav S. <sv...@gm...>]

> On Fri, 2004-08-13 at 08:20, Michael Renner wrote:
> 
> [snip]
> 
> > Do you have information about kernel 2.6 support? And information where
> such 
> > configurations (local-multi-user-X) are used?
> > 

> http://www.mandrakesoft.com/company/press/pr?n=/pr/products/2491
> 
this seems to use Miguel's solution after all 
(hope i didn't broke some trade secrets )

There's also a IT department in Skoda in Chech Republic
and in Izrael a school related project
but no information available on the net

both use several 4-headed boxes

best,

svetljo

-- 
NEU: WLAN-Router für 0,- EUR* - auch für DSL-Wechsler!
GMX DSL = supergünstig & kabellos http://www.gmx.net/de/go/dsl

Re: User questions welcome?

[Andrew C. <an...@di...>]

On Fri, 2004-08-13 at 08:20, Michael Renner wrote:

[snip]

> Do you have information about kernel 2.6 support? And information where such 
> configurations (local-multi-user-X) are used?
> 

http://www.c3sl.ufpr.br/fourhead/index-en.php
http://www.mandrakesoft.com/company/press/pr?n=/pr/products/2491

> CU


Andrew

Re: User questions welcome?

[Aivils <ai...@un...>]

On Friday 13 August 2004 10:20, Michael Renner wrote:
> On Tuesday 10 August 2004 15:44, Aivils wrote:
> 
> [...]
> 
> > First You should try manualy all 6 variants of starting order.
> > startx -- -layout X2 vt17
> 
> thanks for all your information. I got some GeForce II cards this week. With 
> this combination everyting is running fine! 
> Some minor problem like 'shutdown'

X servers must downed in the reverse order of start. gdm do that.
 
> or competing sound access are not to bad. 

Under Mandrake/Fedora exists /etc/security/console.perms
which contains device access permisions
Multiple sound card usage explains only my files
http://www.schuldei.org/aivils/eng-down.html
Check out Mandrake 9.1 files here.
No articles.

> Thanks for your answers
> 
> [...]
> 
> > > I even think about to write an article about this method, so every help
> > > is welcome!
> >
> > Go ahead !
> 
> Do you have information about kernel 2.6 support? And information where such 
> configurations (local-multi-user-X) are used?

2.6 usage equals 2.4 usage.
In theory 2.6 have working fbcon . User may start two framebuffer devices
and each fb device have separate keyboard. Also matrox-dualhead owners can
use each head as seprate text mode console bounded with separate keyboard.
Reports about fbcon is very different. Low level fb device drivers may be broken.
fbcon new features are undocumented and breaks old school  multiheading,
where one user uses number of fbdev heads.
Multiple X servers uses aprox 5% of kernel console code so it is aprox 20 times
more bugless against fbcon.

Aivils

Re: User questions welcome?

[Michael R. <mic...@tu...>]

On Tuesday 10 August 2004 15:44, Aivils wrote:

[...]

> First You should try manualy all 6 variants of starting order.
> startx -- -layout X2 vt17

thanks for all your information. I got some GeForce II cards this week. With 
this combination everyting is running fine! Some minor problem like 
'shutdown' or competing sound access are not to bad.

Thanks for your answers

[...]

> > I even think about to write an article about this method, so every help
> > is welcome!
>
> Go ahead !

Do you have information about kernel 2.6 support? And information where such 
configurations (local-multi-user-X) are used?

CU
-- 
+---------------------------------------------------------+
|Michael Renner                   |                       |
|MPI fuer biologische Kybernetik  |Phone: +49-7071-601-638|
|Spemannstr.38, D-72076 Tuebingen |FAX:   +49-7071-601-616|
|mic...@tu...  |                       |
+----------------------------------------------------ESC:wq

Re: Q: problems with missing /dev/tty0 on X startup in MIPS system

[Pete P. <pp...@mv...>]

>Unfortunately - neither the fbdev code nor the X startup activates the 
>card.
>The X startup failed with an "unable to map the card" once and now 
>repeatedly fails with:
>  
>

I don't know which ATI Rage card you have exactly, but there is a patch 
for the RageXL (tested on a MIPS board) on 
ftp.linux-mips.org:/pub/linux/mips/people/ppopov/2.4/aty_nobiosinit.patch.

The problem is that at kernel version 2.4.24 or .25, don't remember 
which one, the aty code was restructured and the patch does not apply 
anymore.

Pete

Re: Q: problems with missing /dev/tty0 on X startup in MIPS system

[Ralf A. <ra...@KO...>]

Still unable to get the miniPCI ATI Rage to work with the meshcube.

> Yep, that's the current virtual console. Do you have CONFIG_VT enabled? If yes,
> probably the VC initialization failed due to vgacon failing. In that case, you
> can try enabling dummycon (and hope X can wake up your graphics card).
> 
> BTW, there exists (depending on your kernel version) code in atyfb to
> initialize the RAGE XL.

Hello Geert,

many thanks for your help. With the dummycon and the atyfb code activated 
(I'm using a 2.4.24 kernel) I was able to start the X server.

Unfortunately - neither the fbdev code nor the X startup activates the 
card.
The X startup failed with an "unable to map the card" once and now 
repeatedly fails with:

...
(==) Log file: "/var/log/XFree86.0.log", Time: Fri Aug 13 00:05:26 2004
(==) Using config file: "/etc/X11/XF86Config-4"
(EE) No devices detected.

Fatal server error:
no screens found

I do not see whether any int10 module specific code is executed (the X.log 
file does not indicate any module load activities at all.

Any hints here on the list on how to proceed? I'll probably better try 
to ask the Xfree people now.

best regards
 Ralf

Re: Q: problems with missing /dev/tty0 on X startup in MIPS system

[Goswin v. B. <bre...@in...>]

Ralf Ackermann <ra...@KO...> writes:

> Hello,
>
> I'm trying to get VGA output on a MIPSel system (MeshCube 
> http://www.meshcube.org) to work. 
> The system uses a miniPCI ATI Rage VGA card (=> problem does not get 
> initialized due to lack of system BIOS => hopefully initialized by int10 
> XFree86 x86 emulator code).
>
> I've attached an USB mouse and keyboard but fail to start X due to
>
> Fatal server error:
> xf86OpenConsole: Cannot open /dev/tty0 (No such file or directory)

Didn't you see my mail about MAKEDEV?

MfG
        Goswin

Re: Q: problems with missing /dev/tty0 on X startup in MIPS system

[Svetoslav S. <sv...@gm...>]

> Hello,
> 
> I'm trying to get VGA output on a MIPSel system (MeshCube 
> http://www.meshcube.org) to work. 
> The system uses a miniPCI ATI Rage VGA card (=> problem does not get 
> initialized due to lack of system BIOS => hopefully initialized by int10 
> XFree86 x86 emulator code).
> 
> I've attached an USB mouse and keyboard but fail to start X due to
> 
> Fatal server error:
> xf86OpenConsole: Cannot open /dev/tty0 (No such file or directory)
> 
> My questions to those working with MIPS boards / on the Ruby/Linux console
> project are:
>  - 1. How to test the USB input functionality?
> 	(I could test the mouse by doing: cat < /dev/input/mice
> 	=> any hints on doing something similar for the keyboard?)

not a guru here, but i hope this will help you
-----------------------
modprobe evdev
cat /proc/bus/input/devices
: Bus=0011 Vendor=0001 Product=0001 Version=ab41
N: Name="AT Translated Set 2 keyboard"
P: Phys=isa0060/serio0/input0
H: Handlers=kbd event0
B: EV=120003
B: KEY=4 2000000 2b803878 f840d001 f2ffffdf ffefffff ffffffff fffffffe
B: LED=7
<snip>
cat /dev/input/event0
------
should give you some characters on typing to
the corresponding keyboard

>  - 2. Any hints on the missing /dev/tty0 stuff? (this is maybe
> 	related to the console / keyboard stuff?)
> 
> Any hints are welcome, best regards
>  Ralf

best,

svetljo

-- 
NEU: WLAN-Router für 0,- EUR* - auch für DSL-Wechsler!
GMX DSL = supergünstig & kabellos http://www.gmx.net/de/go/dsl

Re: Q: problems with missing /dev/tty0 on X startup in MIPS system

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

On Thu, 12 Aug 2004, Ralf Ackermann wrote:
> I'm trying to get VGA output on a MIPSel system (MeshCube
> http://www.meshcube.org) to work.
> The system uses a miniPCI ATI Rage VGA card (=> problem does not get
> initialized due to lack of system BIOS => hopefully initialized by int10
> XFree86 x86 emulator code).

So virtual consoles are not working.

> I've attached an USB mouse and keyboard but fail to start X due to
>
> Fatal server error:
> xf86OpenConsole: Cannot open /dev/tty0 (No such file or directory)

>  - 2. Any hints on the missing /dev/tty0 stuff? (this is maybe
> 	related to the console / keyboard stuff?)

Yep, that's the current virtual console. Do you have CONFIG_VT enabled? If yes,
probably the VC initialization failed due to vgacon failing. In that case, you
can try enabling dummycon (and hope X can wake up your graphics card).

BTW, there exists (depending on your kernel version) code in atyfb to
initialize the RAGE XL.

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

Q: problems with missing /dev/tty0 on X startup in MIPS system

[Ralf A. <ra...@KO...>]

Hello,

I'm trying to get VGA output on a MIPSel system (MeshCube 
http://www.meshcube.org) to work. 
The system uses a miniPCI ATI Rage VGA card (=> problem does not get 
initialized due to lack of system BIOS => hopefully initialized by int10 
XFree86 x86 emulator code).

I've attached an USB mouse and keyboard but fail to start X due to

Fatal server error:
xf86OpenConsole: Cannot open /dev/tty0 (No such file or directory)

My questions to those working with MIPS boards / on the Ruby/Linux console 
project are:
 - 1. How to test the USB input functionality?
	(I could test the mouse by doing: cat < /dev/input/mice
	=> any hints on doing something similar for the keyboard?)
 - 2. Any hints on the missing /dev/tty0 stuff? (this is maybe
	related to the console / keyboard stuff?)

Any hints are welcome, best regards
 Ralf

---------------------------------------------------------------
Dr. Ralf Ackermann            _         ra...@KO...
Multimedia Communications |/ | | |\/|           Merckstrasse 25
                          |\ |_| |  |  64283 Darmstadt, Germany
Tel.: (+49) 6151 16-6138                Fax: (+49) 6151 16-6152
---------------------------------------------------------------
             http://www.kom.tu-darmstadt.de/~rac
---------------------------------------------------------------

Re: LinuxConsole, yet not working!

[Aivils <ai...@un...>]

> > > > My days are spent configuring and reconfiguring and I have so far
> > > > reconfigured my kernel for some 300 times and installed, yet without
> > > > success.

You have a will!

> > > > I have lately configured as appeared below my signature and it is
> > > > working fine. But I could not see my USB scrollmouse (which works
> > > > perfect without tweaking and in Windows) when I execute the

usb mouse and keyboard cover HID, HID input must figured.
In the 2.4.25 version USB HID is figured in by default

> >
> > # lsmod
> > Module                  Size  Used by    Not tainted
> > iptable_filter          2444   0  (autoclean) (unused)
> > ip_tables              16448   1  [iptable_filter]
> > ppp_synctty             8320   0  (unused)
> > ppp_async               9856   1
> > usbkbd                  4796   0  (unused)
> > uhci                   30492   0  (unused)
> > microcode               7808   0  (autoclean)
> 
> you don't have the driver for usb mouse loaded,
> 
> there are actually two drivers
> - hid.o (which also handles keyboards and is recomended)
> - the older usbmouse.o

usbkbd and usbmouse usage is not recommended by code author Vojtech Pavlik.
$ modprobe usb-uhci
$ modprobe hid 

> > usbkbd.c: :USB HID Boot Protocol keyboard driver

please do not use usbkbd. why?

> > usbdevfs: USBDEVFS_CONTROL failed dev 2 rqt 128 rq 6 len 59 ret -6
> > usb-uhci.c: $Revision: 1.275 $ time 14:23:11 Aug  3 2004
> > usb-uhci.c: High bandwidth mode enabled
> > usb-uhci.c: v1.275:USB Universal Host Controller Interface driver
> 
> > uhci.c: ef40: host controller halted. very bad
> > usb.c: USB disconnect on device 00:1f.2-2 address 2
> 
> this looks like a very serious usb problem

Also exists alternative driver usb-ohci. if usb-uhci halt very oft and
constatly, then try usb-ohci

> > --- demesg-unpluged        2004-08-04 16:46:03.000000000 +0545
> > +++ dmesg-pluged        2004-08-04 16:47:14.000000000 +0545
> > @@ -189,3 +189,5 @@
> >  hub.c: new USB device 00:1f.2-2, assigned address 3
> >  usb.c: USB device 3 (vend/prod 0x4fc/0x5) is not claimed by any active
> > driver. usb.c: USB disconnect on device 00:1f.2-2 address 3
> > +hub.c: new USB device 00:1f.2-2, assigned address 4
> > +usb.c: USB device 4 (vend/prod 0x4fc/0x5) is not claimed by any active
> > driver.
> 
> same adress as the scary warning,
> might be try older/newer kernel
> (but this after trying to load the driver for usb mice)

usb device driver for mouse and keyboard is hid
$ modprobe hid

> 
> > > > Kindly help. I am desperately in need of a
> > > > MultiHead system.
> > > >
> > > > And since the FC2 has adopted the X.org instead of XFree86, will it
> > > > affect to the backstreet ruby?
> > >
> > > it depends on wether you are using the modified Xserver,
> > > or the "echo 1 > /proc/.../video-hack"
> > > in the first case you'll have to patch/recompile xorg
> > > (i've a patch)
> > > in the second no changes are needed

X.org and XFree86 today is very similar.
Also does not extists special "backstreet-ruby" distribution.
I think You should use localy avaible distro. Your FC1 is good.
Youn should not recompile kernel 400 times. If You doubt about
someone, then compile as module and after it will be loaded
automatic by system init scripts or You can load modules
manualy after our sugestion :)
FC1 uses modular kernel and so load modules for You even
if You replace kernel. That works mostly, while module names
stay unchanged. Unknown names are figured in by default,
seems hid is unknown module name for FC1.

Aivils

Re: User questions welcome?

[Aivils <ai...@un...>]

On Tuesday 10 August 2004 13:38, Michael Renner wrote:
> On Tuesday 10 August 2004 10:51, Helge Hafting wrote:
> > Michael Renner wrote:
>=20
> [...]
>=20
> > Feel free to ask, and provide details.  Are you trying to build a
> > multi-console
> > machine right now, or are you planning one?   What exactly are the
> > problems?
>=20
> Moin,
>=20
> I have to operate some 'surf station' at science conferences, so I though=
t=20
> it's a chance to test the backstreet ruby patch with several X servers.
>=20
> I have:
>  patched and compiled a 2.4.25 by my own
>  A debian based Pentium 4 with 1 GB RAM
>  X packages 4.3.0-0pre1v3 from http://www.schuldei.org/debian/bruby
>  1 nVidia AGP card
>  1 nVidia PCI card
>  1 Matrox Millennium
>  2 PS2 keyboards
>  1 USB keyboard
>  3 USB mices
>=20
>=20
> The kernel is started with the parameter 'dumbcon=3D2' by grub
>=20
> I followed the instructions from the XFree-Local-multi-user-HOWTO.txt (so=
=20
> far).
>=20
> Now the problem: As long as there are only the PCI cards are build in=20
> everything is fine! I can use both consoles.
>=20
> Therefore, I have these entries in my /etc/inittab:
>=20
> # For we use "1-16", "17" and "18"
> #X2:23:respawn: su - guest -c "startx -- /usr/X11R6/bin/X2 :2 -layout X2=
=20
> -xf86config XF86Config-4 vt7"
> X0:23:respawn: su - averell -c "sleep 5 ;startx -- /usr/X11R6/bin/X0 :0=20
> -layout X0 -xf86config XF86Config-4 vt17"
> X1:23:respawn: su - joe -c "sleep 10; startx -- /usr/X11R6/bin/X1 :1 -lay=
out=20
> X1 -xf86config XF86Config-4 vt18"

=46irst You should try manualy all 6 variants of starting order.
startx -- -layout X2 vt17
then from xterm start secondary X server and so on.
Also for proper Nvidia initialisation mostly works
X -layout noprefbusid -probeonly
where "noprefbusid" is standard xf86 layout without limitations like
prefbusid. After "probeonly" nvidia kernel module is initalised correct and
user can start/stop xf86 any times.

> But when the AGP card is build in, no matter if I try to start the 'X2' o=
r=20
> not: the machine freeze in the second 'X0' is starting. The last lines in=
 the=20
> logfile are
> (II) Module fb: vendor=3D"The XFree86 Project"
>         compiled for 4.3.0.1, module version =3D 1.0.0
>         ABI class: XFree86 ANSI C Emulation, version 0.2
> (II) Loading sub module "xaa"
> (II) LoadModule: "xaa"
> (II) Loading /usr/X11R6/lib/modules/libxaa.a
>=20
>=20
> Here is the output of a 'lspci'
> cactus:~# lspci
> 00:00.0 Host bridge: Intel Corp. 82845 845 (Brookdale) Chipset Host Bridg=
e=20
> (rev 11)
> 00:01.0 PCI bridge: Intel Corp. 82845 845 (Brookdale) Chipset AGP Bridge =
(rev=20
> 11)
> 00:1d.0 USB Controller: Intel Corp.: Unknown device 24c2 (rev 01)
> 00:1d.1 USB Controller: Intel Corp.: Unknown device 24c4 (rev 01)
> 00:1d.2 USB Controller: Intel Corp.: Unknown device 24c7 (rev 01)
> 00:1d.7 USB Controller: Intel Corp.: Unknown device 24cd (rev 01)
> 00:1e.0 PCI bridge: Intel Corp. 82820 820 (Camino 2) Chipset PCI (rev 81)
> 00:1f.0 ISA bridge: Intel Corp.: Unknown device 24c0 (rev 01)
> 00:1f.1 IDE interface: Intel Corp.: Unknown device 24cb (rev 01)
> 00:1f.3 SMBus: Intel Corp.: Unknown device 24c3 (rev 01)
> 01:00.0 VGA compatible controller: nVidia Corporation: Unknown device 032=
b=20
> (rev a1)
> 02:03.0 Multimedia audio controller: C-Media Electronics Inc CM8738 (rev =
10)
> 02:0a.0 Ethernet controller: Intel Corp. 82557 [Ethernet Pro 100] (rev 0c)
> 02:0b.0 VGA compatible controller: Matrox Graphics, Inc. MGA 2064W=20
> [Millennium] (rev 01)
> 02:0c.0 VGA compatible controller: nVidia Corporation NV11 (GeForce2 MX) =
(rev=20
> a1)

Who knows how compatible is this set of video adapters. Also You have 3 pai=
rs
for tests.

> My XF86Config-4 is attached.

Please add simple layout for probing:
Section "ServerLayout"
=A0 =A0 =A0Identifier =A0 =A0 "noprefbusid"
=A0 =A0 =A0Screen =A0=A0=A0=A0 =A0 =A0 0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 =A0=
=A0=A0=A0=A0=A0=A0"Screen0" 0 0
=A0 =A0 =A0InputDevice =A0 =A0"Mouse0" =A0=A0=A0 =A0=A0=A0=A0=A0=A0=A0"Core=
Pointer"
=A0 =A0 =A0InputDevice =A0 =A0"Keyboard0" =A0 =A0=A0=A0=A0=A0=A0"CoreKeyboa=
rd"
=A0EndSection


> In the documentation the gpm with a special =20
> parameter is mentioned. But I don't use gpm. In an other docu the command
> echo  "1"> /proc/bus/pci/hackvideo
> is mentioned, but as far as I understud the HowTo it's not longer needed =
with=20
> the patched X server?

Patched xf86 does not need for hackvideo. Someone may have unpatched xf86
and will test linux-ruby anyway :) Sometime videoapdapters like prefbusid,
sometime hackvideo. We do not have report statistic.
Patched xf86 does not need for Option        "PciOsConfig" "1"

> I even think about to write an article about this method, so every help i=
s=20
> welcome!
Go ahead !

Aivils