Local "howto" information

[still in the process of updating this. -- rgr, 19-Dec-04.]

Home : Linux resources : "Howto"

This page contains "howto" information for random bits of hardware and software. Some of it is general, but some is very specific to my configuration, so it may break for yours; use caution when trying any of this stuff.

Table of contents

• Automatic clock setting from an NTP time server
• The qmail mail transport agent
• Squid
• Burning CD-R disks
  1. Selected mkisofs and cdrecord options
  2. CD-R how-to
• AOLserver [getting out-of-date].
• Network Information Service (NIS)

Contents of this page:

1. Local "howto" information
   1. Table of contents
      1. Terminology
   2. Networking
      1. Ethernet
      2. General IP networking
      3. If the network doesn't come up after booting
      4. DNS tricks using perl
   3. Servers
      1. Server logs
      2. The ezmlm mailing list manager
      3. netatalk
      4. Samba
   4. Linux guts
      1. Memory management
      2. rpm and rpmfind
      3. XFree86
   5. Hardware
      1. Zip drive
      2. Floppy disk

• Hard drive means just that, though I have a tendency to say "disk" instead. Back when I was introduced to magnetic storage, floppies were quite new, so "disk" without qualification always meant a rigid disk. (Probably one with removable media; you don't see those anymore, because they can't be made very dense.)
• Disk therefore means floppy disk or zip disk.
• CD covers the fourth kind of disk, though I will sometimes use "CD-R disk" when trying to be more specific.

Networking

See also /usr/doc/HOWTO/Ethernet-HOWTO, comp.dcom.lans.ethernet_FAQ and http://www.scyld.com/network/eepro100.html. The /sbin/mii-diag and /sbin/eepro-diag commands (downloaded from the latter) report on detailed Ethernet NIC status.

General IP networking

• The /etc/services file has the port/protocol to service name mapping.
• "netstat -t" reports on open TCP connections.
• Use "cat /proc/net/dev" to get a tabular version of ifconfig output.
• At boot time, the /etc/init.d/paranoid script is started from /etc/init.d/boot.local, then /etc/init.d/network does the boot-time /sbin/ifup calls, and finally /etc/init.d/firewall is called (both of the latter using proper SysV style from /etc/init.d/rc[2345].d/ links). See also the ipchains firewall script page.

1. The host IP address is unknown, so the firewall script can't configure itself.
2. DNS name resolution is unavailable, which makes Squid fail.
3. Because the world is unreachable (not to mention DNS), ntpd can't so much as ask anyone for the time of day.

1. Bring up the interface. Do this via

          # ifup eth0
          

   as root, of course. This may take a while, and it may fail again, especially if a power outage just ended and the DHCP server is swamped with requests (and may still be booting up itself). Do "ifconfig eth0" to verify that eth0 really has an IP address.

2. Bring up the firewall. You will want to do this before starting other services.

          # /etc/init.d/firewall
          

3. Start the time daemon manually.

          # /etc/init.d/xntpd restart
          

   The system clock will get reset almost immediately, which may make the screen go blank, if it moves further ahead than your screen saver delay interval (mine is 10 minutes).

4. Start the remaining services manually.

          # /etc/init.d/squid start
          # /etc/init.d/apache2 start
          

    rgr> perl -e 'use Socket; $iaddr = inet_aton("128.197.54.20"); print join(" ", gethostbyaddr($iaddr, AF_INET)), "\n";'
    DARWIN.BU.EDU  2 4 \200Å6^T
    rgr> 

    % nslookup -q=PTR 19.200.128.24.in-addr.arpa | perl -ne 'print "$1\n" if /in-addr.arpa\s+name = (.*)$/;'
    h0050da615e79.ne.mediaone.net
    % 

Servers

[for suse 8.1, log rotation is now controlled (mostly) by /etc/logrotate.d/syslog, which requires tweaking. -- rgr, 13-May-03.] [squid log rotation is controlled by /etc/logrotate.d/squid, instead of doing "squid -k rotate"; this also needed tweaking. -- rgr, 29-Jun-03.] [need to write up tweaking notes, describe the logrotate package in outline. -- rgr, 5-Jun-04.]

The ezmlm mailing list manager

• The ezmlm home page.
• EZFAQ 0.40 - ezmlm-idx and ezmlm FAQ
• "Reply-To" Munging Considered Harmful: Why mailing lists shouldn't alter "Reply-to:" fields in messages.

• Current development happens at netatalk.sourceforge.net.
• My UG/FAQ is on the http://bmerc-www.bu.edu/needle-doc/netatalk.html page.

In order to get Samba up and running at my site, I found the following resources helpful:

1. SAMBA - opening windows to a wider world (a US mirror of the Samba home site).
2. SMB HOWTO
3. Using Samba (online version of the O'Reilly Using Samba book, November 1999).
4. Windows NT SP3 and 98 Client SAMBA Problem Page
5. LanMan and NT Password Encryption in Samba 2.x [part of the source distribution; I can't seem to find an online version right now. -- rgr, 1-Dec-01.]

Linux guts

• RPM Guide by Eric Foster-Johnson
• Rpmfind.Net
• Finding RPMs (SuSE Wiki)

The "DontZap" option in the "ServerFlags" section of the /etc/X11/XF86Config file disables the Crtl-Alt-Backspace server abort sequence, letting it through to clients. This option is essential for keeping us Zmacs alumni (who expect C-M-DEL to be bound to backward-kill-sexp) from going postal. -- rgr, 19-Mar-00.

Hardware

Mounting:

Use "mount /mnt/zip". If this fails with an error message that looks like this

       mount: the kernel does not recognize /dev/sda4 as a block device
              (maybe `insmod driver'?)

then this is probably because the machine was booted with the Zip drive off. If that's the case, then

       modprobe imm

will probably be sufficient to get /dev/sda4 recognized, after which the drive can be power cycled without needing to repeat the modprobe. (The boot-time modprobe is done in the /etc/rc.d/rc.modules script. -- rgr, 28-Dec-99.)

If /var/log/messages says

       modprobe: can't locate module block-major-8

and lsmod shows imm already loaded, then do

       rmmod imm; modprobe imm

to force reloading. -- rgr, 2-Jul-00.

Unmounting:

"umount /mnt/zip", and then hit the button to eject. Doing "eject /mnt/zip" seems to work, but only for root.

Formatting:

mke2fs /dev/sda4

You can add the "-c" option to check for bad blocks, but I haven't found a bad block yet, and this is what takes most of the time (about 4.5 minutes), as it has to read the entire drive. (My guess is that bad blocks are detected and "repaired" by the drive, so it's no surprise that I never found one. -- rgr, 7-Aug-01.)

Checking for bad blocks:

When doing badblocks on Zip, use 98288 for number of blocks -- this is (- (/ (* 2048 96) 2) 16), for 96 cyl, 2048 blocks/cyl, 0.5k/block, minus some magic factor. If there is some reason to suspect the existence of a bad block, one can also use the "-c" option to e2fsck when checking disk integrity (but see the comment under "Formatting" above).

Checking file system integrity:

Periodically, when mounting Zip disks, the following warning will appear in /var/log/messages:

       EXT2-fs warning: maximal mount count reached, running e2fsck is recommended

To address this, before mounting the disk, do

       e2fsck -f /dev/sda4

The "-f" option will "Force checking even if the file system seems clean" (from the man page); without it, e2fsck does a quick check, so "-f" is more thorough. You can also add the "-c" option to request bad block checking (but see the comment under "Formatting" above).

Erasing:

To erase a Zip disk thoroughly (or any other disk device or partition for that matter), you can do a "write" test to check for bad blocks:

       badblocks -w /dev/sda4 98288

This takes a while, as it writes a number of patterns across the whole disk, and then rereads them. Each read/write pass takes about nine minutes; the whole process is more than half an hour.

Note the existence of the mzip command.

[The "Zip drive backup" section that was here has been renamed and given its own page. -- rgr, 7-Aug-01.]

Floppy disk