Installation of SuSE 6.0 on a PPro 200 computer (05-06-1999)

1. Backed up everything.
2. Boot from boot disk (included)
3. Problem: Couldn't mount CDROM.
4. Solution: From existing Linux system, inserted blank floppy and typed

   dd if=/cdrom/disks/modules of=dev/fd0

   The install program then found the Sound Bastard CDROM at 0x250.
5. Selected "Expert mode", since wished to become an expert.
6. Problem: It immediately became obvious we would need a bigger hard disk.
7. Solution: Ran out and bought a WD AC313000R 13 GB disk.
8. Problem: BIOS recognized disk but only saw the first 8.2493 GB.
9. Solution: The MB was a EPOX PP6-NF, so checked on www.epox.com to find a new flash BIOS. The latest BIOS available was from 12/31/1996. No new BIOS chips were available either. So, tried to use the disk anyway. At first, we connected it to the same IDE connector as the existing disk (ide0), setting jumpers to "slave" and setting the jumpers on our first HD (a ST37450A) to make it a "master". This turned out to be a mistake. Warning: do not do this! Linux fdisk seemed to work, but there were major problems when trying to format the disk. Over the next hour, hundreds of error messages spewed out on the screen, and the system hung at boot-up, requiring a power cycling. This in turn trashed the root partition which had just been installed. This in turn caused considerable gnashing of teeth, and the utterance of more than a few non-technical computer terms.

   Finally, we switched the cable to the second IDE connector, and set the jumpers on both disks to be "single". Even though the BIOS still claimed it was only an 8GB disk, Linux fdisk automatically recognized all 13 GB and almost automatically created 2 partitions of the correct size. Formatting took only 3 minutes. Intallation of SuSE linux proceeded flawlessly after this. The only problem was that it was not possible to change where packages were installed. Thus, it is necessary to select the sizes and mount points for each partition carefully in order to avoid running out of space.

10. Downloading and compiling kernel 2.2.7 (from Sunsite) worked okay, but the new kernel was 507kb and was too big. Thus, we had no choice but to read the manual (/usr/src/linux/README) that came with the kernel.

     cd /usr/src
     rm linux
     ln -s linux-2.2.7 linux                (change to match kernel version)
     cd linux
     cp /boot/bzImage /boot/bzImage.bak     (if present)
     
     make config
     make dep; make clean
     make
     make bzImage
     cp /usr/src/linux/arch/i386/boot/bzImage /boot
     /sbin/lilo          (It should say:  added linux *
                                          added linux-old )
                         (edit /etc/lilo.conf to include old kernel in case
                          of a problem, see below. No need to use rdev
                          if you run lilo.)    

     # LILO Konfigurations-Datei
     # Start LILO global Section
     append="sbpcd=0x250,SoundBlaster"
     boot=/dev/hda
     #compact       # faster, but won't work on all systems.
     read-only
     prompt
     timeout=200
     vga = 6    # force sane state
     # End LILO global section
     # Linux bootable partition config begins
     root = /dev/hda1
     image = /boot/bzImage
             label = linux        # This line starts with tab 
     image = /boot/vmlinuz        # Or /boot/bzImage.bak
             label = linux-old    # This line starts with tab 
     # Linux bootable partition config ends
     # 

1. Problem: XFree86 and MetroX X servers appeared extremely slow, as if they were using SVGA functions. Also, both Xfree86 and MetroX seemed identical in appearance and slow performance.
2. Solution: Installed AcceleratedX. This caused a speedup from 17.3 to 28.5 as measured by tnimage's X Window benchmark.
3. Problem: Telephone in house and modem were making strange noises. Caller ID box on telephone said there were 15 messages from someone named "Data Error".
4. Solution: Change IRQ of com4 so it has a different IRQ from the hard disk. Also, added the following two lines to /sbin/init.d/serial for the Jameco 4 serial port board:

   run_setserial /dev/ttyS2 irq 5 run_setserial /dev/ttyS3 irq 12

5. Problem: Bash prompt is strange.
6. Solution: edit /etc/profile, uncomment PROMPT_COMMAND section where it sets PS1. Changed prompts to:

   "\h:\`pwd -P\`# " (for root) "\h:\`pwd -P\`$ " (for regular user)

   also added "source /etc/mcset" (see below) at end to allow mc to change directories.
7. Problem: Star Office doesn't work, insists on "media key", whatever that is.
8. Solution: Uninstall Star Office. This solved the problem completely; at least the error messages went away.
9. Problem: Infomagic Moo-Tif May 1996 doesn't work (due to glibc incompatibility).
10. Solution: Checked on Infomagic's web site for a new version, only to find a lot of Windows NT stuff, and no glibc version. Purchased a copy of MetroLink Motif Complete instead.
11. Problem: Unable to install MetroLink Motif (install program wouldn't start):

    ./install-motif lib/install.linx86glibc: error in loading shared libraries : undefined symbol: _Xglobal_lock

    (More gnashing of teeth. Invented a couple of new computer terms at this point.)
12. Solution: This was apparently caused by AcceleratedX 4.1.2 overwriting crucial X libraries during its installation. Reformat HD and reinstall OS, setup X for XFree86 (using Sax), then install Motif, then AcceleratedX last. Even after Motif was installed, AcceleratedX version 4 overwrote some critical file, preventing some Motif apps from compiling statically. AcceleratedX version 5 did not do this.
    Moral: Always install Motif before AcceleratedX!!!
    Also, be sure to use only "minimal" installation of AcceleratedX.
13. Problem: Typing Gray Left and Right Arrows produces a '4' or '6' in rxvt.
14. Solution: (This went away by itself after a while).
15. Problem: Networking still had mysterious delays which appeared intermittently. Pinging or telnetting to the linux computer from any other computer sometimes worked instantly, sometimes only after a delay of up to 5.0 minutes (i.e., 300 pings). It was not a problem with reverse hostname lookup, as /etc/hosts of all computers involved contained correct entries for each other, and pinging from other computers or pinging to the IP number still experienced identical delays. It was not a problem with the originating computer, as pinging other hosts worked normally. It also did not seem to be a network problem, as a DOS computer running PC/TCP attached to the same minihub responded instantly to pings. It is not a problem with the Ethernet card, motherboard, or computer, and is not specific to any distribution, as substituting another card, changing the motherboard or computer, or switching from Slackware to SuSE (all for unrelated reasons) had no effect. Switching to another Ethernet port or moving Windows computers to another port had no effect. It could be a problem with driver for 3c509 card - perhaps it falls asleep after several hours of disuse. Interestingly, when one Linux machine had the problem, another one on the same subnet (but on a different port) also often had the same problem. Once network communications were started successfully either by sending something from the affected computer or after extended pings from the remote computer, response was again instantaneous for several hours, but after a few hours of non-use, the network card would fall asleep again. The problem would disappear for weeks at a time, and then suddenly reappear and for several days it would be difficult to login - but only to the Linux computers. Solaris, Irix, and DOS computers were unaffected. This has been a recurring problem ever since the 2.0 kernels came out. Previous identical setups on the same computers, using 1.2 kernels, were fine. Something odd in the network seems to intermittently screw up networking, whether it is RARP or something else, only on Linux machines.
16. Solution: Put

    ping -c 1 {some other host} > /dev/null

    in crontab. This sends 1 ping every half hour to another computer to keep the network card from falling asleep. This solved the problem for a while, but after a month the computer began having delayed responses to ping again (i.e., no response for 1 min) when the ping originated from a ppp connection. When the computer was pinged from another computer on the same subnet, the it responded instantly and simultaneously began responding to the ping over the ppp. Added a line to crontab on this second computer, to ping the first computer every 30 min. This solved the problem (so far).
17. Problem: Middle mouse doesn't paste text in X windows (for AcceleratedX).
18. Solution: In X Windows run /usr/X11R6/bin/Xsetup. Don't use '-graphics' option, this will cause it to crash instead of saving the new configuration. Changed mouse from Microsoft to Logitech Mouseman.
19. Problem: None of the existing X programs on /home would execute.
20. Solution: Recompile tnimage, xscreensaver, xdata, etc. Also recompiled and reinstalled gmp-2.0.2 (math library).
21. Problem: ssh and srp wouldn't compile (ulimit.h not found for ssh, multiple compile errors for srp).
22. Solution: Apparently ulimit(3) is no longer available. Obtained a new version (2.0.12) of ssh. Srp hasn't been updated in 2 years, so it was debugged manually.
23. Problem: None of existing LaTeX documents would compile.
24. Solution: Change from \usepackage{amstex} to \usepackage{amsmath}, and remove all \begin{itemize} and \end{itemize} from around {enumerate} environments. (Can't complain about this, this seems to be an improvement).
25. Problem: Syslogd can't write to xconsole.
26. Solution:

    rm /dev/console ln -s /dev/tty0 /dev/console chown tjnelson.users /dev/tty0 chmod a+rwx /dev/tty0

    This fixed the problem.
27. Problem: Constantly says "Have a lot of fun" at login. Could not find a reason for this error message.
28. Solution: rm /etc/motd. (This did not really solve the problem, still not having much fun yet, but at least the error message went away).
29. Problem: updatedb running at 9:30 pm instead of 3 am. Unable to determine how to change this, as it did not seem to be in crontab or in /var/spool/atjobs, in /etc/, or anywhere else. Also, updatedb not finding any user files, making "locate" useless, even though it was configured to do so in /etc/rc.config. There are 39,514 user files and over 111,000 system files so it is necessary to run updatedb manually every day to repair the database.
30. Solution: Crond was checking the time stamp on some files to determine when to run commands on an hourly, daily, weekly, or monthly basis. Made the following changes:
    1. Edit /etc/rc.config, RUN_UPDATEDB=no
    2. Comment everything out of /etc/crontab. This file called 'run-crons' which is a script that checks the modification times of /etc/cron.daily, etc, then runs a number of scripts based on the results. What a mess. Used "crontab -e" to set up a real crontab. Since the crontab belongs to root, and not using the parameters in /etc/rc.config, it was able to find all the files.
    3. Ran "crontab -e" and added the following:

       # This updates the database for 'locate' every day at 3:40 am: 40 03 * * * updatedb 1> /dev/null 2> /dev/null

       This should automatically refresh the cron daemon, and will show up in the syslog. It also creates a file /var/cron/tabs/root. Read /etc/cron/cron.d, /etc/cron/cron_daily, etc. in case there are other entries to add.
31. Problem: Can't cd to home directory. (UID is different from before)
32. Solution: chown -R tjnelson.users /home/tjnelson
33. Problem: After su'ing, path is quite short. Can't change in /etc/login.defs, /etc/profile, or anywhere.
34. Solution: "strings /bin/su" showed the path to be hard-coded in the su binary. After installing srp, the srp version of su, which replaced the GNU version, functioned correctly. As an aside, it would be nice if srp came with a replacement man page for su as well. The GNU man pages seem to have a lot of ranting and raving.
35. Problem: SuSE manual is useless for finding solutions to installation problems.
36. Solution: Put manual between "Solaris Advanced System Administrator's Guide" and "Internet Security Professional Reference" on bookshelf. (This did not really help; in fact, it seemed to make it worse).
37. Problem: Netscape prints the following message:

    ERROR: libc.so.5: cannot open shared object file: No such file or directory Cant load plugin /usr/X11R6/lib/Acrobat3/Browsers/intellinux/nppdf.so. Ignored.

    Netscape 4.5 crashes when "Preferences..Navigator..Language" is clicked. Netscape 4.5 crashes when it hits a page containing Java (for example, www.novagen.com), taking the X server along with it. This turned out to be a bug in AcceleratedX and was reported to the vendor. Also, 'Filter' in Netscape 4.5 file browser dialog doesn't work properly (just cd's to previous directory no matter what string is entered).
38. Solution: Switch to version 4.6. This required making a link in /usr/lib to satisfy ldd.
39. Problem: Version 4.6 of Netscape crashes continually, whenever one clicks on a menu item, tries to send mail, or views any page containing Java or JavaScript. Netscape forgets its own color settings and reads pages incorrectly if it runs for more than 24 hours.
40. Solution: (This is still not fixed).
41. Problem: Counter on web page doesn't work.
42. Solution: (This is still not fixed).
43. Problem: Can't uninstall packages unless SuSE CD is in drive.
44. Solution: Use rm -rf instead.
45. Problem: Man page for chage(1) annoyingly politically correct.
46. Solution: Edited man page.
47. Problem: Can't print to printer attached to parallel port, even though network printers are all okay.
48. Solution: Parallel port is now /dev/lp0 instead of /dev/lp1. Also, SuSE installed some unknown binary file over /usr/bin/lpf (the printer text filter). Replaced /usr/bin/lpf with backup (see below) and changed entry to lp0 in /etc/printcap.
49. Problem: Odd beeping sound when backspacing on empty line in text mode.
50. Put "setterm -bfreq 0" in /etc/profile.local.
51. Problem: Strange message appearing in log: su[8803]: + ??? root-root This means that root is su'ing to itself on an unknown tty. This also happens on a computer that only has one user and is not connected to a network.
52. Solution: This went away by itself after a while.
53. Problem: /etc/profile and other critical system files being changed by SuSEconfig, destroying several days worth of changes. New files placed in /etc directory are being automatically deleted. Permissions on various files are automatically being changed, which prevents pppd from working. Also, the supath is automatically being changed back and /etc/hosts is being sorted for some unfathomable reason.
54. Solution: copy /sbin/SuSEconfig to a floppy, take floppy out, and dissolve it in 6M sulfuric acid (see software revenge). Restored uncorrupted /etc/hosts from backup.
55. Problem: SCSI scanner recognized at boot-up as sga, but can't open scanner, and no /dev/sga exists; but there is a character device in /dev/sg0 which is 21,0.
56. Solution: Copy /dev/scanner from backup into /dev, then chmod a+rwx /dev/scanner and chmod a+rwx /dev/sg0. This fixed the problem (not sure why). However, if scanner was disconnected, computer must be rebooted before it is found again, just like in Windows. Solution: Made link /dev/sga -> /dev/sg0. This seemed to fix it.
57. Problem: syslogd not saving failed logins, etc. even though its config. file says to do so.
58. Solution: (This is not yet fixed).
59. Problem: anonymous ftp transfers & IP numbers not being logged by syslogd.
60. Solution: Modify ftpd.c, recompile & reinstall. (Can't add command-line options, since srp's ftpd doesn't work with them.)
61. Problem: syslogd date/time stamp is incorrect. Appears to be using the local time of the remote user, so that most ftp logins are recorded as having occurred 4-5 hours into the future.
62. Solution: "Make config" in the kernel did not seem to have an option for disabling this new time warp feature of Linux. Will try a chroniton dampening field.
63. Problem: System demands a change of password on every login, forcing one to type the password 4 times. The /etc/shadow entry looks correct.
64. Solution: Restored the old version of login instead of the srp version.
65. Problem: Clock is always wrong after a boot-up.
66. Solution: (This is not yet fixed). The system clock has never worked correctly in Linux. Even Windows does not have this problem. Why is it so hard for Linux to get the right time?
67. Problem: Meminfo not updating itself every 2 sec as it is supposed to (takes at least a minute after a change before it shows up).
68. Solution: (This is not yet fixed).
69. Problem: fvwm95 ignores its config file, sets background to a sickly green color.
70. Solution: Use tnimage to wallpaper the background with a cement2.gif.
71. Problem: Nedit "String was not found" message box doesn't take Esc key.
72. Solution: Comment out line 349 (where it sets up an 'addEscapeHandler') in DialogF.c and recompile.
73. Problem: Samba clients can't connect or browse network.
74. Solution: 'strings /usr/sbin/smbd' showed that /etc/smb.conf is hard-coded in the smbd binary. Moved smb.conf to /etc. Old version of smbd and nmbd compiled against libc5 failed silently without writing anything to their log files. Type 'ldd smbd' to make sure it is the glibc version, then after starting samba, 'ps -aux | grep mbd' to make sure it is really running. Also, it was necessary to type 'smbpasswd -a [username]' for each user before windows98 could login. Windows 3.11 did not have this problem.
75. Problem: sshd not running.
76. Solution: Added the following to the end of /sbin/init.d/network, after the last esac:

    # Look for sshd in the two most common locations (compiled with --prefix=/usr # or with --prefix=/usr/local) and if we find it, start it up if [ -x /usr/local/sbin/sshd ]; then echo -n " sshd" /usr/local/sbin/sshd elif [ -x /usr/sbin/sshd ]; then echo -n " sshd" /usr/sbin/sshd fi

77. Problem: Lynx crashes, because unable to access its help file.
78. Solution: cp -r /usr/doc/packages/lynx/lynx_help/* from another system.
79. Problem: Netscape 4.6 won't start, libstdc++.so.2.8 => not found.
80. Solution: ln -s libstdc++.so.2.7.2.8 libstdc++.so.2.8
81. Problem: On one computer, SuSE installation program would not install anything, giving a message "Only 80 k left in /mnt/tmp, 200 k needed".
82. Solution: This is caused by the small space allocated for the ramdisk. Reboot and start installation again, loading only the Adaptec AHA1542 driver (which was needed for the CDROM drive on this computer), and not allowing 'autoprobe' to run (which also found and loaded a driver for the 3c509 card).
83. Problem: No applications that use the serial port would compile, gave this error:

    aggregate 'struct termios newtio' has incomplete type and cannot be initialized.

84. Solution: Add #include < asm/termbits.h > to include file and recompile. You would think whoever changed this would have tried to make the serial interface more portable instead of less portable, but nooo.
85. Problem: No applications that use ncurses would compile.
86. Solution: Change #include < ncurses/curses.h> to #include < ncurses.h > and recompile.
87. Problem: Wrong httpd is being run.
88. Solution: Edit /sbin/init.d/apache and /etc/rc.config. Make sure httpd.conf is in /etc/httpd.
89. Problem: Numerous weird users in /etc/passwd.
90. Removed all but essential entries.
91. Problem: 'Last' displays garbage if system is up more than a few days.
92. Solution: (This has not yet been fixed).
93. Each time Netscape opens a second window, between 20 and 40 MB of memory is allocated which is never freed. After about an hour of browsing, all 128 MB of swap space and 128 MB of RAM are completely filled, requiring a reboot to avoid crashing the system.
94. This is not yet fixed.

Back