Comment 4 for bug 8155

Created an attachment (id=200)
more dmesg outputs

Yes, that's correct. When the SATA controller is set to combined mode in the
BIOS (which obviously is necessary to boot from the CDROM), then the CDROM is
not detected by the installer. If I disable the SATA controller, then it's
working fine, but obviously I can't install to the SATA disk in this case.
The CDROM and PATA disk are not on the same IDE bus (this would be tricky with
my current cable setup). It made no difference wheither I had the CDROM on IDE
1 or IDE 2.
The dmesg output I attached earlier was the one from my working Ubuntu
installation, booting from the PATA disk but using the SATA disk as root
filesystem. Meanwhile I also managed to boot Ubuntu from the SATA disk as
described above, as long as the Grub root setting points to the PATA disk (if I
understand this correctly, this loads only the kernel from the PATA disk). This
is probably not an Ubuntu problem anymore and can be ignored for now.

Attached in the tgz (I hope it works, did not want to spam you with four
attachments) are the dmesg outputs of these configurations:
1. Ubuntu booting from my SATA disk with PATA disk attached, CDROM not attached
(I never managed to make my BIOS recognise my PATA disk, the CDROM and the SATA
disk at the same time)
2. Booting from the installation CD with SATA enabled and PATA disk not
attached
3. Booting from the installation CD with SATA disabled and PATA disk attached
4. Booting Fedora from my SATA disk with PATA disk not attached, CDROM attached
(for comparison)

It is interesting to note the difference between booting Ubuntu from the PATA
disk and booting from the SATA disk. In the former case, the IDE driver was
loaded first and created a device hdd for the PATA disk. In the latter case,
the SCSI subsystem was loaded first and a device sdb was created for the disk
(which works just as well).
This pattern is similar for the installation CD. When SATA is enabled, the SCSI
subsystem is loaded first, the CDROM is detected as scsi0, but obviously the
driver can't do anything with it. Later the ide driver does nothing because
"ports already in use". When SATA is disabled, the SCSI part does not happen
and the IDE driver can correctly detect the CDROM and create the device.
Now finally at the Fedora output, it can be seen that Fedora loads the IDE
driver first, the CDROM is correctly detected and loaded and the SCSI subsystem
is loaded after that, ignoring the CDROM drive.
From my ignorant point of view, this looks like the issue is simply that the
IDE driver should always be loaded before the SCSI subsystem.