[CentOS] CentOS4 and older megaraid - SOLVED

[Craig White]

On Sun, 2005-08-21 at 13:06 -0600, William wrote:
> Craig White wrote:
> > this probably isn't the type of task for a Linux newbie - at least not
> > one that gets discouraged easily.
> 
> Exactly; that's why I'm building a driver disk.  Compiling anything, 
> especially against the kernel, isn't for newbies.  That doesn't mean 
> there aren't newbies who need this legacy megaraid in an easy-to-deploy 
> medium.  I'm trying to step in to provide that solution.
> 
> > It does seem that you are completely over thinking the issue of a driver
> > disk. Copy the megaraid.ko to a vfat formatted floppy disk and insert
> > when told to insert driver disk.
> 
> I thought you'd know this, but a driver disk is much more than just a 
> .ko file on a floppy (or several).  I'm not over thinking this, at all. 
>   I'm just challenged by it because this is my first driver disk project 
> and I can't seem to find a full set of detailed requirements for 
> building a successful driver disk (I have provided the rhdd, modinfo, 
> pcitable, modules.cgz, and modules.dep files, but some important detail 
> is still missing).  ddiskit filled in some gaps, but I still don't have 
> the specifications for what would cause, "No matching hardware was found 
> on your system."
> 
> You're right.  I could get my personal server up and running if I 
> followed the steps on the tuxyturvy site, with your additional input, 
> and a 4.1 ISO set.  However, that would help only me and others like you 
> and me who are competent with handling these low-level chores.  I'm 
> dealing with a bigger picture.  I am deliberately trying to build a 
> full-fledged driver disk that will help everyone else.  In reality, our 
> process is very complex to a newbie compared to the alternative: 
> download a driver disk image from some web site and use it with the very 
> convenient 'linux dd' install mode.  I'm trying to provide that driver 
> disk image; one that I can say works with confidence.  That is my only 
> goal.  :)  I'm not at all desperate to get this particular server 
> running.  If I was, I wouldn't be posting on this thread (or the threads 
> that I've started, myself).  :)
> 
> Part of my problem is that this is my hobby, which I only work on over 
> the weekends.  When I started this project, CentOS 4.0 was still 
> available and in wide use.  That said, you're right.  There is no longer 
> any point in trying to continue with 4.0.  I just hope I can make a good 
>   driver disk before 4.1 becomes antiquated.  :)
> 
> After all that, I hope that I have clearly asserted my objective, which 
> isn't quite what you had previously assumed.  :)  You have helped me 
> clarify some important details in the kernel matching requirements (I 
> had no idea i586 and i686 sources couldn't be on the machine at the same 
> time).  Now, if you know where I might find all the specifications and 
> troubleshooting details I need to make a working driver disk (aside from 
> ddiskit), I'm all ears!  :)
-----
but a driver disk for this purpose is of little consequence. 

1. The driver disk would have to be built with the i586 kernel-devel but
when install was completed, it would fail to load the proper module and
fail to boot. There is no way a driver disk would fix this problem
unless this driver disk had all of the appropriate modules in place and
the installer installed each of them to their respective locations and
then ran the mkinitrd.

2. All subsequent kernel upgrades would require re-compile of the kernel
module and rebuild of initrd after the kernel were installed - thus the
driver disk would only further insulate the operator from the
information that they would need to know. In fact, this step would
probably be necessary almost immediately as I would suspect that a
disproportionate amount of users having the LSI hardware raid hba would
also be multi-processor and would need to build for smp kernel.

thus your endeavor to simplify isn't really gonna in the end, help all
that much.

Craig