Initrd

Initial ramdisks, in short initrd, are used in Linux distributions for some time now to keep the core kernel slick and loadable and still all drivers (modules) available to boot on a given hardware.

In contrast to other distributions, such as even including Red Hat Enterprise Linux, the T2 SDE comes with an initrd implementation that ships all IDE, SCSI, ... modules by default that ever could be useful to boot a machine as well as associated automatic hardware detection. This has the big advantage that the underlying hardware is exchangeable without the need to tinker with the initrd - which by design runs in a rather minimal environment to keep the size at a minimum and thus is nothing an end users want to get in touch with.

Additionally the T2 initrd does not contain a static binary, script or Red Hat's commonly found special purpose nash interpreter. Instead it uses a statically linked pdksh shell and real hardware detection utilizing our hotplug++ program. All in all this setup does not only allow a T2 installation to boot without an initrd regeneration or when the underlying system hardware is exchanged (such as on system defect). It also offers a sensible fallback for the user in form of a normal shell - for example when the root filesystem can not be found for whatever reason (competting implementations usually just let the kernel panic in this situation).

Configuration and Regeneration

When a Linux kernel is built by the T2 build system the generic initrd is automatically generated.

To manually regenerate the initrd a script under the commonly known name mkinitd can be used. Without arguments it will generate the image for the currently running kernel. Optionally a specific kernel version can be specified to generate an image for another than the running kernel. The argument -R allows the specification of another directory as system root - usually this is just used by T2 during cross builds, but can also be handy for the use in rescue environments as well.