![[ Prev ]](../gx/navbar/prev.jpg){kind=small}
![[ Table of Contents ]](../gx/navbar/toc.jpg){kind=small}
![[ Front Page ]](../gx/navbar/frontpage.jpg){kind=small}
![[ Talkback ]](../gx/navbar/talkback.jpg){kind=small}
![[ FAQ ]](./../gx/navbar/faq.jpg){kind=small}
![[ Next ]](../gx/navbar/next.jpg){kind=small}
The CD-Writing HOWTO at http://www.guug.de/~winni/linux/and the official cdrecord site at www.fokus.gmd.de/research/cc/glone/employees/joerg.schilling/private/cdrecord.html offer a great deal of information about installing and using CD-ROM writers. Certainly much more than you need to know for an ordinary Linux installation. If you don't want to support an obsolete drive, or an older kernel, or VAX VMS, then you probably don't need to recompile the kernel or make any devices. This short document should be sufficient to get you started burning disks. Once you are started, the official documents will guide you to such esoterica as audio, bootable, multisession and hybrid disks. After determining that RedHat 6.1 was so easy, I tried setting up several other distributions, including RedHat 6.0, SUSE 6.1, Debian 2.1, and Storm 2000. All of those were slightly harder because they didn't include a recent version of cdrecord but none required a kernel rebuild, I have noted the differences along the way. You will need to have:
I believe any drives you bought in a store recently will qualify. The cdrecord docs say that all 1999 or later ATAPI drives support MMC, which is sufficient. Many earlier drives are suppoted also. A look on the shelves at the local computer superstore did not turn up any that mentioned MMC or Linux on the box. My first installation used an older Richoh MP6200A cd recorder. I did more installations with the MagicWriter 4X4X24. This is a very cheap drive, but the manual had a 1999 copyright date and it did work as I expected.
Perform the physical installation of the new drive just as you would any ide drive. It can replace your original read-only drive or be added on. Make sure the drive jumper is set for master or slave as required, the power cable is connected and the data cable has the correct orientation. You shouldn't have to do any CMOS setup. I am told that keeping the cd-writer on a different cable from the hard drive speeds data transfers, but this is probably not significant with Pentium class machines.
At this point stop to check if the BIOS mentions the drive during BIOS initialization. Not all BIOS do that, but especially if it doesn't, you might want to check the cables and the Master/slave jumpers on both the CD-writer and any other drive on the same cable. The /var/log/messages file should have a line about the new "ATAPI" drive.
insmod ide-scsi
RedHat seemed to be the only distribution with an rc.local file. for the others, you have to find some other mechanism for running this command before burning disks and after each reboot.
append="hdc=ide-scsi"
where "hdc" might be "hdb" or "hdd" depending on where your drive is installed. This line should be inserted just after the image statement that boots Linux on your computer. This instructs the kernel to access the cd-writer via the scsi emulation driver.
lilo
cd /cdrom/RedHat/RPMS
rpm --install cdrecord*
rpm --install mkisofs*
The other distributions either included an older version of cdrecord (1.6) which did not support my recorders, or did not include cdrecord at all. If don't have RedHat and need to compile these yourself, the most recent version of cdrecord can easily be found at www.freshmeat.org. I noted that the very extensive cdrecord instructions cover many operating systems, not just Linux (or even Unix). For the distributions considered here, all you need to do is:
tar -xvf cdrecord*tar
cd cdrecord-1.8
make
make install
The cdrecord tarball includes mkisofs. Both packages get installed to /opt/shilly/bin so you will need to make links from a directory in the root path, such as /usr/bin. A very nice feature of cdrecord is that it auto-detects the drive characteristics at run time so there isn't any configuration and you can replace or upgrade the drive without getting into trouble.
/dev/scd0 /cdrom auto defaults,ro,noauto,user,exec 0 0
cdrecord -scanbus
One of the output lines should mention a Removable CD-ROM, and maybe even indicate ``-R'' or ``RW'' to indicate that it is a recorder. Something like this:
0,0,0 0) 'Richoh' 'MP602A' '2.03' Removeable CD-ROM DriveOnly the digit triple at the start is significant. It indicates controller, SCSI ID, and LUN, and is likely to be all zeroes, as shown with SCSI emulation. The only error message I saw in my experiments combined the cases of (1) scsi emulation not correctly installed and (2) drive not found or supported. At that point you might try getting reading to work as an IDE drive in Windows, then in Linux as IDE, then in Linux with SCSI emulation, before concluding that the drive was not supported or broken. I often find this systematic approach to hardware debugging slow but sure.
mkisofs -v -o file.iso file...
where file.iso is your output file and file... is the list of files and directories you want on the cd-rom. If you just list a single directory, the structure is maintained on the CD. Otherwise all the files and subdirectory files are dropped into the root directory with no subdirectory structure. There are a lot of options described in the man page. If you keep your filenames to 8.3 lower case, you won't need to be bothered with most of them. The "-J" option (for Joliet) will allow longer Windows style filenames, but if you actually use longer or case sensitive names your file names will look funny or not work in a minimal ISO9660 system.
cdrecord -v dev=0,0,0 file.iso
The ``dev=0,0,0'' specifies the output device, and might be possibly be different on your system, check the cdrecord -scanbus output if in doubt. Because cdrecord wants to lock pages in memory, it has to be run as root. Making cdrecord setuserid root is endorsed by the Shilly's documentation.
On my 1997 vintage 233mHz AMD with a 5400 rpm hard disk, and a quad speed CD-ROM writer, the system had no trouble maintaining speed and the 512K buffer was never less than 97% full. After initial success you might try combining the mkisofs and cdrecord steps:
mkisofs file... | cdrecord -v dev=0,0,0 -where the hyphen indicates to cdrecord that it should take its input from the standard input. This worked on my system, even when the files were NFS mounted (on a 100BaseTX connection).
Any corrections or suggestions should be sent to me. I am particularly interested in hearing which distributions will work with these minimal instructions and any variations. I don't want to encroach on existing documentation by covering enhanced capabilities - that is well handled already by the existing documentation.
This page is kept at http://www.nber.org/cdrecord.html
[There is another article about CD recording in this issue. -Mike.]