Daniel Gerzo 2006 The FreeBSD Documentation Project &tm-attrib.freebsd; &tm-attrib.general; $FreeBSD$ $FreeBSD$ This document provides information about the NanoBSD tools, which can be used to create &os; system images for embedded applications, suitable for use on a Compact Flash card (or other mass storage medium). NanoBSD NanoBSD is a tool currently developed by &a.phk;. It creates a &os; system image for embedded applications, suitable for use on a Compact Flash card (or other mass storage medium). It can be used to build specialized install images, designed for easy installation and maintenance of systems commonly called computer appliances. Computer appliances have their hardware and software bundled in the product, which means all applications are pre-installed. The appliance is plugged into an existing network and can begin working (almost) immediately. The features of NanoBSD include: Ports and packages work as in &os; — Every single application can be installed and used in a NanoBSD image, the same way as in &os;. No missing functionality — If it is possible to do something with &os;, it is possible to do the same thing with NanoBSD, unless the specific feature or features were explicitly removed from the NanoBSD image when it was created. Everything is read-only at run-time — It is safe to pull the power-plug. There is no necessity to run &man.fsck.8; after a non-graceful shutdown of the system. Easy to build and customize — Making use of just one shell script and one configuration file it is possible to build reduced and customized images satisfying any arbitrary set of requirements. Once the image is present on the medium, it is possible to boot NanoBSD. The mass storage medium is divided into three parts by default: Two image partitions: code#1 and code#2. The configuration file partition, which can be mounted under the /cfg directory at run time. These partitions are normally mounted read-only. The /etc and /var directories are &man.md.4; (malloc) disks. The configuration file partition persists under the /cfg directory. It contains files for /etc directory and is briefly mounted read-only right after the system boot, therefore it is required to copy modified files from /etc back to the /cfg directory if changes are expected to persist after the system restarts. &prompt.root; vi /etc/resolv.conf [...] &prompt.root; mount /cfg &prompt.root; cp /etc/resolv.conf /cfg &prompt.root; umount /cfg The partition containing /cfg should be mounted only at boot time and while overriding the configuration files. Keeping /cfg mounted at all times is not a good idea, especially if the NanoBSD system runs off a mass storage medium that may be adversely affected by a large number of writes to the partition (i.e. when the filesystem syncer flushes data to the system disks). A NanoBSD image is built using a simple nanobsd.sh shell script, which can be found in the /usr/src/tools/tools/nanobsd directory. This script creates an image, which can be copied on the storage medium using the &man.dd.1; utility. The necessary commands to build a NanoBSD image are: &prompt.root; cd /usr/src/tools/tools/nanobsd &prompt.root; sh nanobsd.sh &prompt.root; cd /usr/obj/nanobsd.full &prompt.root; dd if=_.disk.full of=/dev/da0 bs=64k Change the current directory to the base directory of the NanoBSD build script. Start the build process. Change the current directory to the place where the built images are located. Install NanoBSD onto the storage medium. This is probably the most important and most interesting feature of NanoBSD. This is also where you will be spending most of the time when developing with NanoBSD. Invocation of the following command will force the nanobsd.sh to read its configuration from the myconf.nano file located in the current directory: &prompt.root; sh nanobsd.sh -c myconf.nano Customization is done in two ways: Configuration options Custom functions With configuration settings, it is possible to configure options passed to both the buildworld and installworld stages of the NanoBSD build process, as well as internal options passed to the main build process of NanoBSD. Through these options it is possible to cut the system down, so it will fit on as little as 64MB. You can use the configuration options to trim down &os; even more, until it will consists of just the kernel and two or three files in the userland. The configuration file consists of configuration options, which override the default values. The most important directives are: NANO_NAME — Name of build (used to construct the workdir names). NANO_SRC — Path to the source tree used to build the image. NANO_KERNEL — Name of kernel configuration file used to build kernel. CONF_BUILD — Options passed to the buildworld stage of the build. CONF_INSTALL — Options passed to the installworld stage of the build. CONF_WORLD — Options passed to both the buildworld and the installworld stage of the build. FlashDevice — Defines what type of media to use. Check the FlashDevice.sub file for more details. It is possible to fine-tune NanoBSD using shell functions in the configuration file. The following example illustrates the basic model of custom functions: cust_foo () ( echo "bar=baz" > \ ${NANO_WORLDDIR}/etc/foo ) customize_cmd cust_foo A more useful example of a customization function is the following, which changes the default size of the /etc directory from 5MB to 30MB: cust_etc_size () ( cd ${NANO_WORLDDIR}/conf echo 30000 > default/etc/md_size ) customize_cmd cust_etc_size There are a few default pre-defined customization functions ready for use: cust_comconsole — Disables &man.getty.8; on the VGA devices (the /dev/ttyv* device nodes) and enables the use of the COM1 serial port as the system console. cust_allow_ssh_root — Allow root to login via &man.sshd.8;. cust_install_files — Installs files from the nanobsd/Files directory, which contains some useful scripts for system administration. Packages can be added to a NanoBSD image using a custom function. The following function will install all the packages located in /usr/src/tools/tools/nanobsd/packages: install_packages () ( mkdir -p ${NANO_WORLDDIR}/packages cp /usr/src/tools/tools/nanobsd/packages/* ${NANO_WORLDDIR}/packages chroot ${NANO_WORLDDIR} sh -c 'cd packages; pkg_add -v *;cd ..;' rm -rf ${NANO_WORLDDIR}/packages ) customize_cmd install_packages A complete example of a configuration file for building a custom NanoBSD image can be: NANO_NAME=custom NANO_SRC=/usr/src NANO_KERNEL=MYKERNEL NANO_IMAGES=2 CONF_BUILD=' NO_KLDLOAD=YES NO_NETGRAPH=YES NO_PAM=YES ' CONF_INSTALL=' NO_ACPI=YES NO_BLUETOOTH=YES NO_CVS=YES NO_FORTRAN=YES NO_HTML=YES NO_LPR=YES NO_MAN=YES NO_SENDMAIL=YES NO_SHAREDOCS=YES NO_EXAMPLES=YES NO_INSTALLLIB=YES NO_CALENDAR=YES NO_MISC=YES NO_SHARE=YES ' CONF_WORLD=' NO_BIND=YES NO_MODULES=YES NO_KERBEROS=YES NO_GAMES=YES NO_RESCUE=YES NO_LOCALES=YES NO_SYSCONS=YES NO_INFO=YES ' FlashDevice SanDisk 1G cust_nobeastie() ( touch ${NANO_WORLDDIR}/boot/loader.conf echo "beastie_disable=\"YES\"" >> ${NANO_WORLDDIR}/boot/loader.conf ) customize_cmd cust_comconsole customize_cmd cust_install_files customize_cmd cust_allow_ssh_root customize_cmd cust_nobeastie The update process of NanoBSD is relatively simple: Build a new NanoBSD image, as usual. Upload the new image into an unused partition of a running NanoBSD appliance. The most important difference of this step from the initial NanoBSD installation is that now instead of using the _.disk.full file (which contains an image of the entire disk), the _.disk.image image is installed (which contains an image of a single system partition). Reboot, and start the system from the newly installed partition. If all goes well, the upgrade is finished. If anything goes wrong, reboot back into the previous partition (which contains the old, working image), to restore system functionality as fast as possible. Fix any problems of the new build, and repeat the process. To install new image onto the running NanoBSD system, it is possible to use either the updatep1 or updatep2 script located in the /root directory, depending from which partition is running the current system. According to which services are available on host serving new NanoBSD image and what type of transfer is preferred, it is possible to examine one of these three ways: If the transfer speed is in first place, use this example: &prompt.root; ftp myhost get _.disk.image "| sh updatep1" If a secure transfer is preferred, consider using this example: &prompt.root; ssh myhost cat _.disk.image.gz | zcat | sh updatep1 Try this example if the remote host is not running neither &man.ftpd.8; or &man.sshd.8; service: At first, open a TCP listener on host serving the image and make it send the image to client: myhost&prompt.root; nc -l 2222 < _.disk.image Make sure that the used port is not blocked to receive incoming connections from NanoBSD host by firewall. Connect to the host serving new image and execute updatep1 script: &prompt.root; nc myhost 2222 | sh updatep1