Dan Carpenter wrote:
On 8/4/06, john <fedora@xxxxxxxxxx> wrote:
? I can build and run the DVD no problem but when the install
completes, the
system can't boot from the HDD because the module is not there to be
found.
Any suggestions welcome.
Driver disks did work with rh7.3.
The fact that the install completes means that the driver disk is at
least working a bit. Otherwise you wouldn't be able to install to the
3ware array.
What kind of a system are you installing to, Intel or AMD? What's the
output from zcat modules.cgz | cpio -t?
regards,
dan carpenter
Thank you, .... thank you for your reply !!
To answer your question first, its a vanilla P4 Intel system. But I
think you may be misunderstanding me a little (or perhaps I am
misinterpreting your answer!).
Its not that driver disks don't work... its that the kickstart
'driverdisk' option doesn't work.
Before explaining this more clearly, I should point out that the DVD is
only able to access the 3ware RAID array onto which it installs a new
OS because I have specially prepared a custom cdboot.img that contains a
3ware driver module that is compatible with the boot kernel used by the
DVD. I can explain how I did this if you like but suffice to say all
this works fine. The DVD has no trouble reading the array and correctly
installs everything onto the HDD.
The problem I have relates to how anaconda handles the drivers destined
for use by the kernel that is booted from the HDD. At the risk of
telling you what you already know, let me explain this a bit more (sorry
if its a bit tedious).
If I prepare a 3ware driver diskette and install the OS normally by
inserting this diskette into the drive when prompted by the installer,
everything installs correctly. When I subsequently boot off the hard
disk, the newly installed system (smp kernel) boots perfectly. This is
because anaconda ensures that once the driver diskette has been read,
it's contents are preserved until near the end of the install, at which
point it rebuilds an initrd for each kernel supported by the diskette
(up, smp, bigmem etc) and copies the result into the HDD boot partition.
Once the DVD installation has completed and system is rebooted, the new
OS and its supporting initrd are loaded from the HDD's boot partition
and the drivers required to access the RAID array are ready and waiting.
Does this make sense ? If so, its worth emphasizing that the 3ware
drivers loaded via diskette are NOT used by the kernel on the DVD
because its module has already been provided via the DVD's own initrd
image. (Phew!)
Now, back to my problem. All I'm trying to do is remove the need to
insert the driver diskette by placing a copy of the driver disk's files
(module-info, modules.dep, modules.cgz pcitable) onto the DVD itself and
then telling anaconda to go looking for my drivers on the DVD rather
than prompting me for a diskette.
The 7.3 manual says I can do this using the line
driverdisk hda --type=iso9660
in my ks.cfg kickstart file.
This is apparently supposed to be interpreted as 'go and look in the
root of the hda partition and you will find some drivers that you need
to load'. But it isn't. Unfortunately, what happens is that anaconda
doesn't appear to look where it's told, it therefore doesn't keep track
of my special 3ware drivers and in the end, doesn't rebuild the initrd
image needed by the smp kernel that gets loaded when I finally boot
drectly from the HDD.
Is this any clearer... or have I made things worse !?!!
Regards,
John
