The XINU Page

4 months ago 4

Experimenting With Xinu

There are several recent versions of Xinu available for an x86 architecture hardware board, an ARM architecture hardware board, and Oracle's (free) VirtualBox hypervisor.

In our lab, we use two computers for Xinu: a conventional computer used to compile a Xinu image, and second,, otherwise idle, back-end computer, into which we download and run the image. The second edition of the text contains code for both the Intel x86 (Galileo) and ARM Cortex 8 (BeagleBone Black), and the code is freely available. Either of the platforms is relatively inexpensive and available from commercial web sites.

There are several ways to experiment. For the BeagleBone Black, the quickest was to get started involves using a serial cable. Connect the serial cable from the BBB to a PC, download a Xinu image over the serial cable, and then enter a command that causes the BBB to run the image. You can download a instructions that explain the above in detail for the BeagleBone Black; there are two sets:

For the Galileo, the quickest way to get started involves using an SD memory card and a serial cable. Connect the serial cable from the board to a PC (inexpensive cables are all that one needs). Run a terminal emulator, such as minicom to display the serial output in a window on the PC. Plug the SD card into the PC, compile Xinu and load the image onto the SD card. Unplug the SD card from the PC and plug it into the experimenter board. Boot the board, and use the console serial connection to boot the image from the SD card. Output from Xinu will be displayed over the serial line.

The second way to experiment takes more time to set up, but allows one to recompile and download quickly once the pieces are in place. As described above, a serial line must be connected from the experimenter board to a PC. In addition, one must: connect a board to an Ethernet switch, connect a PC to the same switch, configure the board to boot over the network, and configure the PC to run DHCP and TFTP server processes that supply the information needed to boot. Once all the above has been set up, it is possible to compile Xinu, place the image in a file on the PC, and then power-cycle the experimenter board to boot the Xinu image automatically. We use this approach in the lab to enable a rapid edit-compile-download-test cycle. You can download a document that explains the above in detail for the Galileo board (both Generations 1 and 2).

We enthusiastically endorse the two approaches above because they allow Xinu to run on real hardware and provide users with the best experience in designing systems. For situations in which users cannot afford an experimenter board or prefer not to touch real hardware, we have a version of Xinu that runs in a Virtual Machine (VM) environment, which means it can run on a conventional computer with no extra hardware. The Virtual Box version is based on code from the second edition of the text, but has a few differences. For example, because the emulated console uses port-mapped I/O instead of memory mapped I/O, the fetch and store operations on CSRs in the Xinu tty driver have been replaced by calls to macros that either generate port-mapped or memory mapped refernces (the macros are defined in io.h). Because the emulated Ethernet differs from the Galileo Ethernet, the Ethernet driver has been replaced. Finally, references to other memory-mapped functions have been replaced (e.g., the low-level PCI bus functions have been reverted to use the generic x86 version).