Re: Matrox G550 mga driver hangs system

[Tony Stoneley]

Felix Miata wrote on Fri, 31 Mar 2017 18:17:09 -0400
>...So, a bug filing is apparently needed. I'd like you to file it

OK, with great trepidation I've sent a report to the Debian Bug
Tracking System, and received automatic acknowledgement of it as
Bug#859423. Those interested should follow progress via
www.debian.org/Bugs

Tony

Re: Matrox G550 + mga driver hangs system

[kamaraju kusumanchi]

On Wed, Mar 29, 2017 at 9:36 AM, Tony Stoneley  wrote:
>
> A severely trimmed version of /var/log/Xorg.0.log follows. The whole
> affair is 616 lines, which seems excessive for an initial posting.
> This is with the fully up-to-date testing distribution, in particular
> with xserver-xorg-video-mga version 1:1.6.5-1.

In Xorg.0.log, the errors are logged with (EE) prefix, so the output of

grep EE /var/log/Xorg.0.log

would be useful in documenting the problem. In your case, you already
gave the relevant information but thought I would point it out in case
you are not aware of it.

raju
-- 
Kamaraju S Kusumanchi | http://raju.shoutwiki.com/wiki/Blog

Re: Matrox G550 mga driver hangs system

[Felix Miata]

I uninstalled tdm-trinity, kept iomem=relaxed on cmdline, blacklisted mga, and 
rebuilt initrd. That got Xorg to run TDE and IcwWM via startx. Then I 
reinstalled tdm-trinity, and its greeter now behaves as expected.

--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/

Re: Matrox G550 mga driver hangs system

[Felix Miata]

Tony Stoneley composed on 2017-03-31 22:20 (UTC+0100):


Felix Miata wrote on Fri, 31 Mar 2017 02:22:10 -0400



It may be time for you to ask for help from the devs, using the
debian-devel mailing list or one of the freedesktop.org Xorg mailing
lists, or by filing a Debian bug.



Yes, perhaps, if I can pluck up enough courage...


The problem has returned here. Xorg.0.log keeps terminating with an empty 
backtrace, and nothing in dmesg unless adding drm.debug= to cmdline. C-A-D 
doesn't do anything. openSUSE Tumbleweed still works. Nothing I can think of has 
changed overnight. Maybe this is come kind of timing issue, consistently bad for 
your machine, occasionally OK for mine.


Last night and this AM before my last post here I was reliably getting into Xorg 
after a lot of trial & lockups. So, a bug filing is apparently needed. I'd like 
you to file it. You're using XFCE4, so presumably only standard repos, while I'm 
using Trinity, meaning non-standard repos, plus having a related issue you 
haven't mentioned. TDM is trying to start even though the target is multi-user, 
so ATM at least I can't get a chance to even try startx.


Once you've filed and shared a bug I'll subscribe and add comments. Bug filing 
instructions I just used for an installation problem are here:

https://www.debian.org/releases/stretch/i386/ch05s04.html.en

Main instructions seem to be different:
https://www.debian.org/Bugs/Reporting
--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/

Re: Matrox G550 mga driver hangs system

[Tony Stoneley]

Felix Miata wrote on Fri, 31 Mar 2017 02:22:10 -0400
  

>Your goal is to boot without Plymouth and without framebuffer, in
>80x25 mode, to give Xorg the best possible chance to work as expect.
>If Plymouth is installed, purge it.
>
>To proceed, hit the e key when the Grub menu appears, then remove any
>line that says "load_video", and from any line that includes "video="
>or "vesa" or "vga=" or "quiet" or "splash", remove each whole such
>string. Optionally, if the string "text" appears, remove it too.
>
>If all the above doesn't help, repeat it, but append "iomem=relaxed" to the
>line
>that included video and/or vesa and/or vga.

OK, I did all that. More strictly I edited grub.cfg (I do have a
rescue CD, so not a disaster if I broke grub.cfg) to comment out two
calls of load_video, one in the menu entry for the boot and one in the
preamble. The grub dialogue duly came up in 80x25 mode, encouragingly,
but even with "iomem=relaxed" on the "linux" line I still had no joy
when it came to starting X, alas.

>Are you using a greeter, or logging in on a vtty and using startx or
>equivalent?

For this investigation I am using the latter, but normally I use
lightdm and its greeter. I dispensed with lightdm here to simplify
while trying to sort out what was going on.

>What are the permissions on your /usr/bin/Xorg?

-rwxr-xr-x


>Doing something like 'journalctl -b -1 | grep -i failed' might be
>useful. There is an awful lot of stuff making particular points of
>interest hard to identify in the journal.

It certainly is hard to find the needle in the haystack, even when as
in my case much of it is gobbledegook. I've tried the suggestion but
it didn't throw up anything very startling.

One thing I have found is that although ctrl-alt-Fn has no observable
effect when the xserver has jammed, ctrl-alt-del does provoke reboot,
and the shutdown is reasonably orderly, as seen later in journalctl,
so the underlying system is still flying.

Thanks for all the other helpful instruction, even if it hasn't solved
the problem.

>It may be time for you to ask for help from the devs, using the
>debian-devel mailing list or one of the freedesktop.org Xorg mailing
>lists, or by filing a Debian bug.

Yes, perhaps, if I can pluck up enough courage...

Tony

Re: Matrox G550 mga driver hangs system

[Felix Miata]

Felix Miata composed on 2017-03-31 02:22 (UTC-0400):

[xserver-xorg-video-mga is a user-space graphics driver]
...

If all the above doesn't help, repeat it, but append "iomem=relaxed" to the line
that included video and/or vesa and/or vga...

This showed up on vtty when I did dist-upgrade from jessie to stretch:

apt-listchanges: News
...
linux-latest (76) unstable; urgency=medium

  * From Linux 4.8, several changes have been made in the kernel
configuration to 'harden' the system, i.e. to mitigate security bugs.
Some changes may cause legitimate applications to fail, and can be
reverted by run-time configuration:
- On most architectures, the /dev/mem device can no longer be used to
  access devices that also have a kernel driver.  This breaks dosemu
  and some old user-space graphics drivers.  To allow this, set the
  kernel parameter: iomem=relaxed

xorg-server (2:1.17.3-1) unstable; urgency=medium

  The Xorg server is no longer setuid root by default.  This change reduces the
  risk of privilege escalation due to X server bugs, but has some side effects:

  * it relies on logind and libpam-systemd
  * it relies on a kernel video driver (so the userspace component doesn't
touch the hardware directly)
  * it needs X to run on the virtual console (VT) it was started from
  * it changes the location for storing the Xorg log from /var/log/ to
~/.local/share/xorg/

  On systems where those are not available, the new xserver-xorg-legacy package
  is needed to allow X to run with elevated privileges. See the
  Xwrapper.conf(5) manual page for configuration details.
--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/

Re: Matrox G550 mga driver hangs system

[Felix Miata]

Tony Stoneley composed on 2017-03-30 17:12 (UTC+0100):


Felix Miata wrote on Wed, 29 Mar 2017 16:44:01 -0400



https://lists.debian.org/debian-user/2017/03/msg00895.html



Did you try other things suggested in that thread or the openSUSE bug
referenced there
https://bugzilla.opensuse.org/show_bug.cgi?id=1004453 ?



To be honest, I'm way out of my depth in all that. I would be happy to
try a particular experiment, given instructions...



e.g. disabling framebuffer?



Er, how? xserver-xorg-video-fbdev isn't installed, nor for that matter
xserver-xorg-video-modesetting. As I said, I'm out of my depth.


A framebuffer is historically how the boot messages get displayed on vtty1 by 
the kernel in other than 80x25 text mode. Plymouth can be and often is used to 
convert that process to a graphical mode. By disabling framebuffer I mean to 
ensure that you are booting in 80x25 mode. Matrox doesn't support all the usual 
standard VESA modes, so getting back to the most basic video output can be key 
to video problem solutions.


Your goal is to boot without Plymouth and without framebuffer, in 80x25 mode, to 
give Xorg the best possible chance to work as expect. If Plymouth is installed, 
purge it.


To proceed, hit the e key when the Grub menu appears, then remove any line that 
says "load_video", and from any line that includes "video=" or "vesa" or "vga=" 
or "quiet" or "splash", remove each whole such string. Optionally, if the string 
"text" appears, remove it too.


If all the above doesn't help, repeat it, but append "iomem=relaxed" to the line 
that included video and/or vesa and/or vga. If this works, and Grub2 is what you 
are using, then /etc/default/grub needs to be modified to match whatever worked, 
followed by running update-grub. If still using Grub, simply update menu.lst to 
match what worked.



Which is yours PCIe, or AGP?



Ah! One I can answer:  AGP


That's what I have.


Which WM/DE(s) is/are you trying to use?



xfce4 and all that goes with it, but I don't think it's getting that
far. As previously remarked, that stuff does all work with the vesa
driver (achieved by tweaking xorg.conf).


FBDEV and VESA Xorg drivers are creepy-crawly slow!!!

Are you using a greeter, or logging in on a vtty and using startx or equivalent?

What are the permissions on your /usr/bin/Xorg?


Can you see any other clues than Xorg.0.log shows by running
'journalctl -b -1'?



Nope, though I might possibly not recognise a clue...


Doing something like 'journalctl -b -1 | grep -i failed' might be useful. There 
is an awful lot of stuff making particular points of interest hard to identify 
in the journal.



Apologies for uselessness (and also btw for wrecking the thread
structure with a completely inadvertent small subject change, the
genesis of which is a complete mystery to me).


Don't be confused by the fact that some common video terms have multiple 
contexts. Using the VESA driver in Xorg has nothing directly to do with VESA 
modes being used by the kernel or the BIOS. Same goes for modesetting or KMS. 
Matrox gfxchips are not supported by KMS, so the modesetting Xorg driver is not 
an option for Matrox users. xserver-xorg-video-modesetting is appropriate only 
for Intel, ATI and NVidia hardware that is several years newer than your Matrox. 
xserver-xorg-video-fbdev would probably not work as well as xserver-xorg-video-vesa.


Lack of active Matrox support since KMS was introduced into the kernel around 8 
years ago is why were are going through this troubleshooting process. Devs are 
no longer Matrox users, so must rely on users who have problems reporting them 
with enough details that fixes can be implemented by devs who have no matching 
hardware to test on. It may be time for you to ask for help from the devs, using 
the debian-devel mailing list or one of the freedesktop.org Xorg mailing lists, 
or by filing a Debian bug.

--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/

Re: Matrox G550 mga driver hangs system

[Tony Stoneley]

Felix Miata wrote on Wed, 29 Mar 2017 16:44:01 -0400
>> https://lists.debian.org/debian-user/2017/03/msg00895.html

>Did you try other things suggested in that thread or the openSUSE bug
>referenced there
>https://bugzilla.opensuse.org/show_bug.cgi?id=1004453 ?

To be honest, I'm way out of my depth in all that. I would be happy to
try a particular experiment, given instructions...

>e.g. disabling framebuffer?

Er, how? xserver-xorg-video-fbdev isn't installed, nor for that matter
xserver-xorg-video-modesetting. As I said, I'm out of my depth.

>Which is yours PCIe, or AGP?

Ah! One I can answer:  AGP

>Which WM/DE(s) is/are you trying to use?

xfce4 and all that goes with it, but I don't think it's getting that
far. As previously remarked, that stuff does all work with the vesa
driver (achieved by tweaking xorg.conf).

>Can you see any other clues than Xorg.0.log shows by running
>'journalctl -b -1'?

Nope, though I might possibly not recognise a clue...

Apologies for uselessness (and also btw for wrecking the thread
structure with a completely inadvertent small subject change, the
genesis of which is a complete mystery to me).

--
Tony Stoneley

Re: Matrox G550 + mga driver hangs system

[Felix Miata]

I tried to net install Stretch to my MGA550 machine, but kept getting segfaults 
trying to configure network. So I restored a backup image of Jessie from another 
machine to my MGA550 machine. It worked normally, so I dist-upgraded it to 
Stretch on vtty 3. When done, I logged into :1 using startx with a good TDE 
session. Then I exited and logged in in the greeter. That worked too. I logged 
out, then tried to log back in, and got a black screen and no keyboard response.


I rebooted, and tried various cmdline options and switching back and forth 
between graphical and multi-user targets for over an hour. Sometimes I'd get 
black screen and no keyboard response. Sometimes X would start and immediately 
exit. Sometimes a segfault would show up in dmesg. Sometimes a segfault would 
show up in Xorg.0.log. I had a bug almost ready to file before figuring out 
using multi-user.target how to reliably get a working TDE session using this 
kernel cmdline:

root=LABEL=SS25deb9 ipv6.disable=1 net.ifnames=0 noresume vga=791 iomem=relaxed

I'm actually starting via a script, /usr/local/bin/tdestart:
#!/bin/sh
WINDOWMANAGER=/opt/trinity/bin/starttde startx

xserver-xorg-legacy is installed, and /usr/bin/Xorg perms are set to 4711.

Adding drm.debug=1 or drm.debug=0x0e to cmdline causes black screen, no KBD 
response, and the following in dmesg & the journal:

kernel: mtrr: no MTRR for f400,200 found
kernel: [drm] Initialized
kernel: [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
kernel: [drm] No driver support for vblank timestamp query.
kernel: [drm] Initialized mga 3.2.1 20051102 for :01:00.0 on minor 0
--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/

Re: Matrox G550 mga driver hangs system

[Felix Miata]

Tony Stoneley composed on 2017-03-29 20:28 (UTC+0100):


Felix Miata suggested
https://lists.debian.org/debian-user/2017/03/msg00895.html



Thanks, Felix, but I had already seen that and tried it, to no great
effect. I should have said so. Except that I see the log I filed
doesn't say so. Here is (part of) one that does, and fails similarly-



[  6651.505] X Protocol Version 11, Revision 0
[  6651.505] Build Operating System: Linux 3.16.0-4-amd64 i686 Debian
[  6651.505] Current Operating System: Linux tresco 4.9.0-2-686-pae #1
SMP Debian 4.9.13-1 (2017-02-27) i686 [  6651.505] Kernel command line:
BOOT_IMAGE=/boot/vmlinuz-4.9.0-2-686-pae
root=UUID=920046b5-b826-4c18-85a7-c986f9c15dfd ro quiet iomem=relaxed
[  6651.505] Build Date: 03 March 2017  03:15:35PM [  6651.506]
xorg-server 2:1.19.2-1 (https://www.debian.org/support)
 ...
[  6659.399] (EE)
[  6659.399] (EE) Backtrace:



Apologies. Problem remains, however.


Did you try other things suggested in that thread or the openSUSE bug referenced 
there

https://bugzilla.opensuse.org/show_bug.cgi?id=1004453 ?
e.g. disabling framebuffer?

Note in that bug I declared the problem solved using the 4.10 kernel, but 4.9 
works too.


Which is yours PCIe, or AGP?

Which WM/DE(s) is/are you trying to use?

Can you see any other clues than Xorg.0.log shows by running
'journalctl -b -1'?
--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/

Re: Matrox G550 + mga driver hangs system

[Tony Stoneley]

to...@tuxteam.de wrote on Wed, 29 Mar 2017 16:29:52 +0200
>is the system totally dead or just the display? Can you ping (or even
>ssh) your box? In case you have no network, the effort of setting one
>up (locally) might well pay off.

I wish I could, but alas I only have one PC here, albeit on a small
LAN, connected to the world by a portmapping link. A new machine is on
my wish list. Maybe I will have to bring Christmas forward, or lean on
a kindly friend to loan a machine...

As to the underlying question, I think the system is still somewhat
alive, since the disc active light flickers now and then.

Tony

Re: Matrox G550 mga driver hangs system

[Tony Stoneley]

Felix Miata suggested
https://lists.debian.org/debian-user/2017/03/msg00895.html

Thanks, Felix, but I had already seen that and tried it, to no great
effect. I should have said so. Except that I see the log I filed
doesn't say so. Here is (part of) one that does, and fails similarly-

[  6651.505] X Protocol Version 11, Revision 0
[  6651.505] Build Operating System: Linux 3.16.0-4-amd64 i686 Debian
[  6651.505] Current Operating System: Linux tresco 4.9.0-2-686-pae #1
SMP Debian 4.9.13-1 (2017-02-27) i686 [  6651.505] Kernel command line:
BOOT_IMAGE=/boot/vmlinuz-4.9.0-2-686-pae
root=UUID=920046b5-b826-4c18-85a7-c986f9c15dfd ro quiet iomem=relaxed
[  6651.505] Build Date: 03 March 2017  03:15:35PM [  6651.506]
xorg-server 2:1.19.2-1 (https://www.debian.org/support)
 ...
[  6659.399] (EE) 
[  6659.399] (EE) Backtrace:

Apologies. Problem remains, however.

-- 
Tony Stoneley, a...@ajms.ukfsn.org

Re: Matrox G550 + mga driver hangs system

[Felix Miata]

Tony Stoneley composed on 2017-03-29 14:36 (UTC+0100):


I'm trying an experimental upgrade from stable to testing on an
elderly system with a Matrox Millenium G550 graphics adaptor plugged
directly into the motherboard, using the mga driver, and have run into
a brick wall with the xserver. When started, eg with startx, the
system immediately freezes and the only escape seems to be reboot. In
particular ctrl-alt-Fn has no effect in this state, which makes
diagnosis tricky. Any help or suggestions would be appreciated.



[Mandatory declaration, having read the rubrics: I am not subscribed
to this list (for fear of drowning). I just read the web archive


Seriously?
Date: Thu, 23 Mar 2017 11:11:52 -0400
Subject: Re: Matrox G550 - MGA X server fails
https://lists.debian.org/debian-user/2017/03/msg00895.html
--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

 Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata  ***  http://fm.no-ip.com/

Re: Matrox G550 + mga driver hangs system

[tomas]

-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1

On Wed, Mar 29, 2017 at 02:36:55PM +0100, Tony Stoneley wrote:
> I'm trying an experimental upgrade from stable to testing on an
> elderly system with a Matrox Millenium G550 graphics adaptor plugged
> directly into the motherboard, using the mga driver, and have run into
> a brick wall with the xserver. When started, eg with startx, the
> system immediately freezes and the only escape seems to be reboot. In
> particular ctrl-alt-Fn has no effect in this state, which makes
> diagnosis tricky. Any help or suggestions would be appreciated.

Sorry I can't help you much with X, but perhaps one remark/question
which might be of interest to those who can:

is the system totally dead or just the display? Can you ping (or even
ssh) your box? In case you have no network, the effort of setting one
up (locally) might well pay off.

> [Mandatory declaration, having read the rubrics: I am not subscribed
> to this list (for fear of drowning). I just read the web archive. I
> post this with some trepidation...]

No worries :-)

> Perhaps there was more that never got flushed to disk, but who knows.
> More of the elided material available on request of course, together
> with any other info.

Another thing you could clear up if you can access your box through
the back door.

Regards
- -- tomás
-BEGIN PGP SIGNATURE-
Version: GnuPG v1.4.12 (GNU/Linux)

iEYEARECAAYFAljbxOAACgkQBcgs9XrR2kbSDQCdGuzy1LMJNZaiYjOERJqBwnR2
9HIAnRP5PvDpfH0FGoWV9wEPaV7yxzVx
=wMcd
-END PGP SIGNATURE-