NanoBSD is designed to put a possibly reduced FreeBSD system on a Compact Flash card (or other mass storage of your choice) in a way which is suitable for use in appliance like applications.
The mass storage is divided into three parts, two image partitions and a configuration file partition, which normally are mounted only read-only. This means that it is safe to pull the power plug on a NanoBSD machine and that Flash based storage is not worn out with filesystem metadata writes.
Having two image partitions means that it is possible to download a new image while the system is running, reboot to run that new image and if that fails somehow, it is still possible to switch back to the old image partition.
The Nanobsd build system lives in src/tools/tools/nanobsd and consists of makefiles and shell scripts which drives use of the Makefiles in the FreeBSD source tree to create a NanoBSD image.
The necessary commands to build a NanoBSD image are:
Depending on your machine, this will take from half an hour to some hours. The process is split into a sequence of steps, and if errors happen underway it is generally, but not always possible to rectify the error and type make again to complete the missing steps.
If you want to start from scratch you have to delete all files and directories in the NanoBSD directory which start with an underscore: rm -rf _.*
Here is a breakdown of the files and directories which NanoBSD builds, in the order they are built:
The output from running make buildworld in the source tree.
If this step fails, the output is instead left in the file _.bw.tmp
The output from running make installworld DESTDIR=_.w in the source tree.
If this step fails, the output is instead left in the file _.iw.tmp
This is the directory tree into which installworld installs.
The output from running make distribute DESTDIR=_.w in the source trees etc subdirectory.
If this step fails, the output is instead left in the file _.di.tmp
The output from running make buildkernel in the source tree.
If this step fails, the output is instead left in the file _.bk.tmp
The output from running make installkernel DESTDIR=_.w in the source tree.
If this step fails, the output is instead left in the file _.ik.tmp
The output from running the customization scripts in the Customize subdirectory.
If this step fails, the output is instead left in the file _.cs.tmp
The output from building the image file containing the filesystem images of the finished NanoBSD.
If this step fails, the output is instead left in the file _.md.tmp
Mtree(8) output from the files installed into the code patitions in the NanoBSD image.
Image file containing the full disk image. This is the image file you will put on the Compact Flash or other mass storage from which your target platform boots.
Image file containing just the code partition image. This is image you will download to update a running system.
This is where things get interesting, and where you will be spending your time when developing with NanoBSD.
The first thing you need to configure is how much storage you have on your mass storage media. NanoBSD can work out of the box with as little as 64MB, and if you spend time on it, you can cut it down even further until you end up with an installed system with just the kernel and two or three files in userland.
You can determine the size of the storage media with the diskinfo(8) command or by examining dmesg(8) or console output during boot.
Once you know the number of sectors on the entire media insert this in the Makefile in the definitions of SECTS or pass it as a command line option to make(1) when you build NanoBSD: MAKE SECTS=100000.
Some platforms need to use CHS addressing in the bootblocks in which case you need to set the HD and SC values in the Makefile as well. These can also be found from diskinfo(8), dmesg(8) or console output. Even if your current platform does not need these to be set, it is always a good idea to do it anyway.
Please be aware that you have to get the values used on the target platform. Putting a CF card in a USB-CF adapter will not result in the correct values for the geometry.
The DATASIZE parameter controls how many sectors are used for the configuration partition. Typically you will need very little, but depending on your application and how you organize things, you may want to set this higher or lower. Obviously, the space you use for the configuration partition is not available for the image partitions.
You can run scripts which will customize the NanoBSD image by listing them in the CUSTOMIZE variable in the Makefile. The scripts will be given arguments to show them where everything is and they can do pretty much anything they want to do to the directory tree.
The make.conf file overrides the usual /etc/make.conf and by default disables a lot of FreeBSD from the build in order to make it fit on 64MB devices. Go through the list and decide if there are anything you want or do not want in your NanoBSD image.
You may also want to use a kernel other than GENERIC. Put the config file in ../../../sys/i386/conf and set the name in KERNCONF in make.conf.
You can save a couple of megabytes of space in your image by removing all the bits you will not need. (How many appliances need support for nine different RAID controllers ?)
- By default the follwing customize scripts are provided (but not + By default the following customize scripts are provided (but not called unless you list them in CUSTOMIZE.
Disables getty(8) on the VGA devices and enables console in the COM1 serial port.
Disables the "beastie-menu" in the bootloader.
Disables ata-dma for the Soekris NET4801 because early hardware versions do not support DMA on the CF socket. &footer;