[TriLUG] confusion regarding need for initrd during boot

[rpjday]

  i was poking around with roy vestal's problem, seeing if i could
reproduce it, and ended up confused about whether i understand
the need for an initrd ram disk during boot.

  as many of you know, the whole point of building an initrd
file with "mkinitrd" is to make available to the kernel essential
modules that it needs to boot (SCSI, IDE, and the like).  in fact,
the initrd file that ships with red hat contains the ext3.o
module, to support mounting of ext3 filesystems in case you
decide to format using ext3 at install time.

  fair enough.  so i started with a fresh red hat 2.4.7-10 RPM,
built it using their template i686 config file (which selects
ext3 support as a *module*), created the modules, did the basic
install, ran "mkinitrd" to build an initrd image in /boot,
and added the appropriate stanza to /boot/grub/grub.conf.
not surprisingly, it booted just fine.

  just for laughs, i removed the "initrd" line from grub.conf to
see what would happen if the kernel had no pointer to an initrd
file from which to load an ext3 module.  still booted just fine.
wha..???  

  so i explicitly deleted that initrd file from /boot, just in
case the kernel was being clever and looking for it there anyway.
still booted just fine.

  ok, so what's going on?  can ext3 filesystems be mounted
even without an available ext3 module or built-in support for
ext3 in the kernel?  what gives?

rday