Armed with a CD containing Red Hat Linux 5.2, courtesy
of the excellent PC Plus magazine
(May 1999), I was ready to start planning my installation.
I realised that I had two choices when installing Linux
alongside Windows - to re-partition the disk and re-install Windows and
Linux in the new partitions, or to create partitions on the live
Windows system. I decided to do the latter, since the thought of
re-installing Windows 98 was not a pleasant one.
Again I faced two choices - the installation
instructions in PC Plus described a disk partitioning tool being
included as part of Linux. I had also learnt about a highly recommended
product PartitionMagic from PowerQuest.
I decided to use PartitionMagic, on the grounds that it was available
in a Windows-friendly form. I was comfortable with the idea of creating
disk partitions within Windows first, then attempting the Linux
installation as a second stage.
PartitionMagic was also said to include a boot manager
process called BootMagic, with a point-and-click interface for boot
choices. My wife, who shares the system, prefered to be hidden from
anything to do with Linux. It seemed better to install BootMagic, in
preference to the Linux Lilo software, and configure it to make Windows
the default.
With these decisions made and a rough plan of attack, I
set forth to install PartitionMagic, BootMagic and finally Linux. The
actual installations were all remarkably smooth, and are not described
in detail.
X Window System
Having installed the system and logged in as root, one
of my first jobs was to check out the configuration of the X Window
System. I typed startx and was pleased to see the familiar diagonal
cross mid-screen, followed by the Linux desktop (which I gather is
called Gnome in Red Hat Linux). There was a start menu, Control Panel
and a single terminal window. I could use the start menu to create new
terminal windows. The only problem I encountered at this early stage
was - the screen seemed to be shifted off the right edge of my monitor.
During installation, I had selected the option "HF SVGA
1024x768@70Hz". For no very good reason, other than a suspicion that I
might be pushing the limits of my monitor, I decided to try another
option - "Non-interlaced SVGA 1024x768@60Hz". Re-running startx was now
much more successful - I saw the clock and various menus bottom right
of my monitor. Aware of the complexities of the X Window System (I
managed Unisoft's project to produce an X11R5 test suite for MIT
between 1990-1992) I decided not to "probe" further (ha ha).
Printer configuration
During installation, I pretended my HP Deskjet 895Cxi
printer was a generic HP Deskjet. I now attempted to print a simple
text file using the lpr command. The file emerged, but the printer
settings had failed to map the linefeed characters to linefeed-carriage
return on output. As a result, it had the "stair-stepping" effect
mentioned as an option in the printer installation notes. Using the
printer configuration tool accessed from the Control Panel, and setting
the appropriate option, the problem was quickly solved. During the
course of this experiment, I changed the printer type to HP Deskjet 550
- which so far seems to have made absolutely no difference to anything.
Many years ago, I was involved in the development of a
printer control package at ROOT Computers (later Unisoft). I remember
the standard UNIX printcap(5) database included an attribute which
could be used to specify that printers did not map linefeed on output
(and so the printer software emulated this). I assumed that if I looked
in the file /etc/printcap that I would see an attribute (named CR or
something). I looked but did not see anything of the sort. The only
change made by the configuration tool seemed to be to change the {}
characters to 1. Most odd. Obviously the warnings about only editing
the file if you know what you are doing are very apt.
Mounting Windows file systems
I was interested to know whether I could mount the
Windows file system on device /dev/hda1 from within Linux. I had been
told this was possible, but given no specific instructions.
I decided to start in a small way and mount Windows
read-only. I created a new empty directory to act as the mount point:
mkdir /windows
My first attempt
mount -r /dev/hda1 /windows
succeeded immediately. I could change to the /windows
directory and lo and behold, DOS files appeared before my very eyes.
I then became somewhat more confident and tried to mount
the device read-writeable. The command
mount /dev/hda1 /windows
also had the desired effect. Now I am no longer obliged
to use Frontpage Express or notepad if I don't want to. I am editing
this file using vi though it resides in C:. Neat.
I decided to make the Windows file system mount
automatically on reboot. I toyed with the idea of putting an entry in
/etc/mnttab, until I found that there is no such file on my version of
Linux. I chanced on a note in the March 1999 issue of PC Plus,
describing the file /etc/fstab. How could I have forgotton that, I
wondered. From the manual page for fstab(5), I realised I should
probably have been using the mount option "-t vfat" for Windows. I
tried the command
mount -t vfat /dev/hda1 /windows
and found the file system mounted with subtle
differences. Spaces in Windows file names were now apparent; before
they had not been. This made for some fun trying to change into the
directory "My Documents" ! Finally to ensure the desired mount on
reboot, I added the line
/dev/hda1 /windows vfat defaults 0 0
into /etc/fstab, exactly as recommended by PC Plus, and
everything worked fine.
I discovered that the vi editor within Red Hat Linux
understands that Windows files with carriage returns on every line are
reported as being in DOS format, and the carriage returns are
preserved. In order to remove them from a file, you can use the vi
command ":set fileformat unix". When the file is written, it is then
written in UNIX format (i.e. without carriage returns). If a file has
carriage returns on only some of the lines, the file is NOT reported as
a DOS format file, and the carriage returns are explicitly displayed as
^M.
I also discovered that Windows file systems created on
floppy devices can be mounted easily under Linux. For example, the
following command mounted a floppy formatted under Windows and makes
the Windows files appear in the directory /mnt/floppy:
mount -t vfat /dev/fd0 /mnt/floppy
Similarly the following command can be used to make the
Red Hat Linux files from the original Windows CD appear in the
directory /mnt/cdrom:
mount /dev/cdrom /mnt/cdrom
I could then change to the directory
/mnt/cdrom/doc/rhmanual/manual and view the Red Hat installation guide
from within Linux using the command:
netscape index.html
Some months later, I noticed many of these issues of
Windows and Linux co-existence were covered in the July 2000 issue of
PC Plus.
Linux Internet connection
In the months after I installed Red Hat Linux 5.2,
further articles in PC Plus magazine
discussed the stages involved in connecting a Linux system to the
Internet.
In December 1998, the utility EzPPP was mentioned. This
claimed to simplify the configuration of dial-up connections to just a
few simple steps - such as specifying the phone number, user name and
password used with an existing Internet connection. This sounded
promising. Apparently EzPPP included the "chat" program which also
mentioned in the May 1999 issue. The article mentioned different
protocols for sending packets from my PC to the Internet (PPP, SLIP,
and so on). Which one to use ? I rebooted Windows, and found amongst
the Control panel settings evidence that PPP was used, so, I decided to
try that first.
Establishing PPP connections
The next problem was finding EzPPP. It was not on the
December 1998 Super CD as claimed. I found it on the January 1999 Super
CD instead. Better late than never. Also, the GUI front end was not
found - I found just a single source file eznet.c, together with a
document which included installation and usage instructions.
I compiled the source file using gcc as specified and
installed the eznet binary file in /usr/bin, making setuid root as
required. I ran the command I hoped would configure my ISP (Global
Internet):
eznet add service=global user=cater password=*****
phone=08450807777
(Of course I'm not telling you my password, stupid!)
Then I bravely tried the command I hoped would dial in
and create an Internet connection:
eznet up global
This failed with a usage error. Prompted by the
documentation, I looked in the directory /var/eznet and found various
log files. One of these included error messages from pppd (which eznet
attempts to start). These refered to a problem with the first argument
to pppd, the device file /dev/modem. I checked - and found this did not
exist. I knew the COM1 port on Windows was used for the modem. So I
created a symlink to /dev/ttyS0 (as mentioned in the June 2000 PC
Plus). I then saw different error messages from pppd - claiming that
kernel support for PPP was absent. This was serious - I had visions of
compiling and installing new kernels.
Check the documentation
I consulted the document PPP-HOWTO in doc/HOWTO on the
Red Hat distribution. I also came across the README files in the
directory /usr/doc/ppp-2.3.5. In the end I was led to examine the
syslog file for any evidence of kernel messages which refer to PPP. I
first had to look in /etc/syslog.conf to work out that my syslog file
is located in /var/log/messages. I was glad I looked. I found that the
kernel had indeed dynamically loaded the PPP module when I ran eznet.
As a desperate last try, I removed the file /dev/modem
and added an entry to the file /var/eznet/eznet.conf specifying tty as
ttyS0. Running eznet again - to my amazement - started my modem
dialling and created what looked like a connection!
In the course of this I looked more at the log files in
/var/eznet, and ran the eznet log command. This seemed to just
concatenate various of the log files. The log output includes a full
transcript of the login conversation with the ISP, which would be
useful in case of problems. There were also messages logged by pppd
outlining the PPP connection set up.
Nameservers
I ran netscape in another terminal window and entered a
URL. More error messages occurred. These were connected with something
I had read about but chosen to ignore until now - DNS entries. I had
not set up a nameserver. Not surprisingly, there was no way to resolve
www.reuters.com into the actual IP address for my browser to connect
to. In the Red Hat installation manual I discovered sections previously
skipped over concerning setup of Network connections. Soon I came
across the section on nameservers. Although the screens listed were for
gnome-linuxconf (which I don't seem to have installed yet) the
equivalent screens were easily found in the Control Panel (under
Network Configuration->Names). I added the IP addresses of the two
nameservers listed at the Global Internet site. I saved the results,
and out of curiosity looked in the file /etc/resolv.conf, mentioned in
several of the articles on PPP. To my delight, the file had been
created, and the contents looked believable. I restarted netscape,
entered www.reuters.com - and was immediately connected to the
Internet. Soon I was checking my email account, bookmarking sites in
Netscape, and so on. Connected at last.
|