Re: Linux 2.6.8.1 + megaraid-2.20.3.0 == Oops

["Randy.Dunlap" <rddunlap@osdl.org>]

On 21 Aug 2004 13:00:21 -0400 Patrick J. LoPresti wrote:

| "Randy.Dunlap" <rddunlap@osdl.org> writes:
| 
| > Looks like mraid_mm_init() hasn't done the
| > 	INIT_LIST_HEAD(&adapters_list_g);
| > yet.  You might be able to verify that by using
| > "initcall_debug" on the kernel boot/command line.
| 
| Oh, I see now.  Yes, the initicall_debug trace shows that
| mraid_mm_init() is not being called.  But megaraid_init() is, which
| ultimately leads to the crash.
| 
| Is there some way to tell Linux about this dependency when the drivers
| are non-modular?

First (simplest), try reversing the order in scsi/megaraid/Makefile.
That may work.  (put megaraid_mm.o before megaraid_mbox.o)
If that isn't sufficient, it will probably have to be done with initcall
levels (in place of module_init).  See include/linux/init.h for details.
Example:  use:

device_initcall(mraid_mm_init);
and
late_initcall(megaraid_init);



and earlier I wrote:
> I expected to be able to build these drivers as modules, but that's
> not allowed by their Kconfig.  Is that because of the module
> inter-dependencies?

and Pat replied:
| I do not understand.  The first thing I did was to build these as
| modules against the Fedora Core 2 kernel, but that resulted in a
| different failure.  So I thought I would try them non-modular on a
| vanilla kernel, which is how I ended up with this Oops.

The megaraid Kconfig file does not allow these drivers to be built
as loadable modules.  I was asking why not.

--
~Randy