Re: Crashing X servers

[scorpio dodds net]

	To test this, I opened 16 windows.  Admittedly, most of them are
Eterms, so this probably can't be called a valid test, but I'm not really
sure what to test.  Everything seems to be working fine for me; I'm using
an sx164 w/64MB of ram, modified RH5.0 install.  Matrox Mil. II, 4MB.
Hardware cursor.

	I'm using an X server I just recently grabbed from
ftp://data.microway.com/.  Get the new XF86_SVGA (not XF86_SVGA_matrox;
that's compiled for old libs, e.g., w/RH4.2), and give it a whirl.  I
can't guarantee it'll work, but it's worth a shot.

	Oh, and I'm running a 2.1.115 kernel.  You didn't mention which
kernel you're using; if this is a kernel problem, that's quite relevant.
(you said 'factory installed RH5.1'... did the factory update the kernel?
;-)

	If the new X server doesn't fix the problem, run the following
script and send me (or the list, I 'spose) the output.  Also, I'll need to
know the kernel version, and the memory violation message that gets logged
(i.e., the one you sent us this time; it may be different numbers then).

#!/bin/sh
ldd /usr/X11R6/bin/XF86_SVGA
echo
echo /proc/ksyms
cat /proc/ksyms

# -- end of script

On Mon, 31 Aug 1998, Mark Hansel wrote:

> How many people are having X86_SVGA crash when "several" (10-15 somewhere) 
> windows are open, even with only 8 bit modes on? This is a reproducible on
> my LX / 128MB. The crash kills everything.  The keyboard is dead, mouse
> registers nothing, telnet, even ping from another machine get no
> response. 
> 
> At the recommendation of DCG tech support, I added sw_cursor to
> XF86_CONFIG and commented out all but 8bit display modes. No joy.
> 
> Do other people have this problem? How widespread a problem is it? I've
> been told it is a "known" problem. Has anyone found an X server for the
> Matrox Millenium that works? Can we get this thoroughly documented? 
> 
> System description and my practices: Linux is Red Hat 5.1, factory
> installed, XF86_SVGA compile date 5/5/98 and I have tried one later
> version (8 something).  I start xdm from a root command line and the
> machine returns to that command line with a blinking cursor on the crash.
> 
> The kernel does manage to write the following in /var/log/messages: 
> 
> Aug 31 19:31:18 alpha kernel: X: memory violation at pc=12004fc40
> rp=12004fed0 (bad address = 206cbc04)
> 
> This is a BAD and intollerable bug. I do not have the expertise to fix it. 
> I'm a social scientist, not a kernel hacker -- the experienced but humble
> user whom the Linux community is could attract. The Linux kernel is not
> supposed to crash! Several years using Slackware on intel has led me to
> expect stable performance for months at a time (power outage to power
> outage), even when running very buggy programs (mine). 
> 
> I have work to do. I need the power of this machine (and at the end of
> development, much more) to do that work. If there is no fix real soon, I
> have to find a stable OS. Can we get this solved?
> 
> Mark Hansel (218-236-2039)
> hansel@mhd1.moorhead.msus.edu
> 
> 
> -- 
> To unsubscribe: send e-mail to axp-list-request@redhat.com with
> 'unsubscribe' as the subject.  Do not send it to axp-list@redhat.com
>