linuxconsole-dev

HoHo

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

What version of X are you running?
Have you or anybody else have positive experience with
Zephyr. I have tried Ander's binary patched Xephyr
with your 2.6.14-vz2 kernel and xorg 6.9 and get an
I/O error, whereas it works OK for Ander.
YBG
> Hoho
> 
> Two new 2.6.14 kernels at
> http://members.westnet.com.au/vanzeeland :
> 
> vz3.3 : 
> Built from redhat errata 1.1653 FC4 kernel +
> 2.6.14.4 + Faketty 0.04 + Hijack LED + Bluetooth
-mh2 +
> vesafb-tng-1.0-rc1-r2-2.6.14 + fbsplash +
> realtime LSM + shfs + MS PPP and CHAP2
> authentication support.
> 
> vz3.3.1 :
> adds Con colivas ck7 patches
> 
> Good news : with these kernels and nV 7676 or 8xxx
> the P4 that I cursed so much in previous discussion
is now showing no sign of
> text/framebuffer console corruption, or any problem
with X and console switching.

Re: HoHo

[James v. Z. <ja...@dv...>]

re: xorg (patched with isolate device) : 6.8.2-37 (FC4 P4)/ 6.7.0-14
(FC2 SMP Athlon), nVidia 8xxx binary driver, 2.6.14-vz3.3xx kernel looks
like a majic combination for me so far.

re: Zephyr; Not yet - I will probably look at it soon - I have SVHS
outputs that I want to take advantage of for PVR and editing use; I am
cautious on nested solutions - it strikes me, and I may be wrong, as a
high overhead solution, no matter which way you turn. If zephyr is
stable and reliable and takes up some of the discussed features, then
I'm sure to start using it, but it probably won't be for multi-console
use, generally speaking, although does raise the potential for reliable
use of multi-head cards where this is not currently feasable. (including
SVHS outputs :-)

The issues specifically resolved with these vz3.3xx kernel are : any
console switching instability - I've tried quite hard and can no longer
reproduce it. Even on a machine for which this has historically been a
problem.

vesafb-tng is better, but not 100% - on SMP and hyperthreaded machines,
no problem. On SMP boxes it remains unreliable.

I use vesafb-tng console on the HT P4 and vga16 framebuffer console on
the SMP server. Neither show any sign of the issue. I haven't tested
much with text mode console.

On the other hand, you do want to 
> retain <ctrl><alt><bkspc> to reset the card. 

Absolutely. As Aivils pointed out - it's an emergency out. What I'm
wondering is if there would be any benefit to introduce a forced timeout
inserted between <ctrl><alt><bkspc> killing the server and the new X
server attempting to init video. Does a single head fail similarly?

With multi-console at least, using this as a stress test will cause a
crash - either of the X server or a lockup or similar. It is attempting
to re-init immediately. How might one insert a pause of say 0.5 or 1
second, to ensure that this is not the whole problem? Is this
unnecessary?

Of course it's not a real-world usage and it's not normally a problem,
but it would be nice to know what causes it and perhaps prevent it
altogether.

J

On Sun, 2005-12-25 at 16:17, Yitzhak Bar Geva wrote: 
> What version of X are you running?
> Have you or anybody else have positive experience with
> Zephyr. I have tried Ander's binary patched Xephyr
> with your 2.6.14-vz2 kernel and xorg 6.9 and get an
> I/O error, whereas it works OK for Ander.
> YBG
> > Hoho
> > 
> > Two new 2.6.14 kernels at
> > http://members.westnet.com.au/vanzeeland :
> > 
> > vz3.3 : 
> > Built from redhat errata 1.1653 FC4 kernel +
> > 2.6.14.4 + Faketty 0.04 + Hijack LED + Bluetooth
> -mh2 +
> > vesafb-tng-1.0-rc1-r2-2.6.14 + fbsplash +
> > realtime LSM + shfs + MS PPP and CHAP2
> > authentication support.
> > 
> > vz3.3.1 :
> > adds Con colivas ck7 patches
> > 
> > Good news : with these kernels and nV 7676 or 8xxx
> > the P4 that I cursed so much in previous discussion
> is now showing no sign of
> > text/framebuffer console corruption, or any problem
> with X and console switching.
> 
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
> for problems?  Stop!  Download the new AJAX search engine that makes
> searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
> http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
> _______________________________________________
> Linuxconsole-dev mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linuxconsole-dev
> 

Re: HoHo

[James v. Z. <ja...@dv...>]

on SMP and hyperthreaded machines,

should read on uniprocessor and hyperthreaded

On Sun, 2005-12-25 at 23:06, James van Zeeland wrote:
> re: xorg (patched with isolate device) : 6.8.2-37 (FC4 P4)/ 6.7.0-14
> (FC2 SMP Athlon), nVidia 8xxx binary driver, 2.6.14-vz3.3xx kernel looks
> like a majic combination for me so far.
> 
> re: Zephyr; Not yet - I will probably look at it soon - I have SVHS
> outputs that I want to take advantage of for PVR and editing use; I am
> cautious on nested solutions - it strikes me, and I may be wrong, as a
> high overhead solution, no matter which way you turn. If zephyr is
> stable and reliable and takes up some of the discussed features, then
> I'm sure to start using it, but it probably won't be for multi-console
> use, generally speaking, although does raise the potential for reliable
> use of multi-head cards where this is not currently feasable. (including
> SVHS outputs :-)
> 
> The issues specifically resolved with these vz3.3xx kernel are : any
> console switching instability - I've tried quite hard and can no longer
> reproduce it. Even on a machine for which this has historically been a
> problem.
> 
> vesafb-tng is better, but not 100% - on SMP and hyperthreaded machines,
> no problem. On SMP boxes it remains unreliable.
> 
> I use vesafb-tng console on the HT P4 and vga16 framebuffer console on
> the SMP server. Neither show any sign of the issue. I haven't tested
> much with text mode console.
> 
> On the other hand, you do want to 
> > retain <ctrl><alt><bkspc> to reset the card. 
> 
> Absolutely. As Aivils pointed out - it's an emergency out. What I'm
> wondering is if there would be any benefit to introduce a forced timeout
> inserted between <ctrl><alt><bkspc> killing the server and the new X
> server attempting to init video. Does a single head fail similarly?
> 
> With multi-console at least, using this as a stress test will cause a
> crash - either of the X server or a lockup or similar. It is attempting
> to re-init immediately. How might one insert a pause of say 0.5 or 1
> second, to ensure that this is not the whole problem? Is this
> unnecessary?
> 
> Of course it's not a real-world usage and it's not normally a problem,
> but it would be nice to know what causes it and perhaps prevent it
> altogether.
> 
> J
> 
> On Sun, 2005-12-25 at 16:17, Yitzhak Bar Geva wrote: 
> > What version of X are you running?
> > Have you or anybody else have positive experience with
> > Zephyr. I have tried Ander's binary patched Xephyr
> > with your 2.6.14-vz2 kernel and xorg 6.9 and get an
> > I/O error, whereas it works OK for Ander.
> > YBG
> > > Hoho
> > > 
> > > Two new 2.6.14 kernels at
> > > http://members.westnet.com.au/vanzeeland :
> > > 
> > > vz3.3 : 
> > > Built from redhat errata 1.1653 FC4 kernel +
> > > 2.6.14.4 + Faketty 0.04 + Hijack LED + Bluetooth
> > -mh2 +
> > > vesafb-tng-1.0-rc1-r2-2.6.14 + fbsplash +
> > > realtime LSM + shfs + MS PPP and CHAP2
> > > authentication support.
> > > 
> > > vz3.3.1 :
> > > adds Con colivas ck7 patches
> > > 
> > > Good news : with these kernels and nV 7676 or 8xxx
> > > the P4 that I cursed so much in previous discussion
> > is now showing no sign of
> > > text/framebuffer console corruption, or any problem
> > with X and console switching.
> > 
> > 
> > 
> > -------------------------------------------------------
> > This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
> > for problems?  Stop!  Download the new AJAX search engine that makes
> > searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
> > http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
> > _______________________________________________
> > Linuxconsole-dev mailing list
> > Lin...@li...
> > https://lists.sourceforge.net/lists/listinfo/linuxconsole-dev
> > 
> 
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
> for problems?  Stop!  Download the new AJAX search engine that makes
> searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
> http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
> _______________________________________________
> Linuxconsole-dev mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linuxconsole-dev
> 

No luck with vz3-3 or with Aivils patch

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

I tried James' patch with 2.6.14 (really fast!), but
still getting crashes after repeated logouts and
logins (beween 6 and 10). I also recompiled X with
Aivils' patch. After repeated logouts, the machine
freezes. I have no clue what causes the freeze. With
Aivils' patch, an occasional X server hits 99% cpu
usage with an interesting phenomenon: it slows
performance on servers :1 and :2, but not on :0. Only
:0 is not USB.
We also see that each logout takes longer and longer
until the machine freezes.
Yitzhak

Re: No luck with vz3-3 or with Aivils patch

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

On Otrdiena, 27. Decembris 2005 07:22, Yitzhak Bar Geva wrote:
> I tried James' patch with 2.6.14 (really fast!), but
> still getting crashes after repeated logouts and
> logins (beween 6 and 10). I also recompiled X with
> Aivils' patch. After repeated logouts, the machine
> freezes. I have no clue what causes the freeze. With
> Aivils' patch, an occasional X server hits 99% cpu
> usage with an interesting phenomenon: it slows
> performance on servers :1 and :2, but not on :0. Only
>
> :0 is not USB.
>
> We also see that each logout takes longer and longer
> until the machine freezes.

Oops. My patch is wrong. It works for kdm. I confuse
gdm and kdm. gdm don't like my patch :-(
And gdm kills X anyway after 10 sec time out.
I am so sorry.

Aivils

Re: HoHo

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

On Sv=E7tdiena, 25. Decembris 2005 15:06, James van Zeeland wrote:
> With multi-console at least, using this as a stress test will cause a
> crash - either of the X server or a lockup or similar. It is attempting
> to re-init immediately. How might one insert a pause of say 0.5 or 1
> second, to ensure that this is not the whole problem? Is this
> unnecessary?

I think at least gdm tried start up X 3 times with 5 sec time out.
You can wrap X, start wrapper from a DM:
#!/bin/bash
sleep 1
exec Xorg $*

Aivils