The LFS system will be built by using an already installed Linux distribution (such as Debian, OpenMandriva, Fedora, or openSUSE). This existing Linux system (the host) will be used as a starting point to provide necessary programs, including a compiler, linker, and shell, to build the new system. Select the “development” option during the distribution installation to be able to access these tools.

As an alternative to installing a separate distribution onto your machine, you may wish to use a LiveCD from a commercial distribution.

Chapter 2 of this book describes how to create a new Linux native partition and file system. This is the place where the new LFS system will be compiled and installed. Chapter 3 explains which packages and patches need to be downloaded to build an LFS system and how to store them on the new file system. Chapter 4 discusses the setup of an appropriate working environment. Please read Chapter 4 carefully as it explains several important issues you need be aware of before beginning to work your way through Chapter 5 and beyond.

Chapter 5 explains the installation of a number of packages that will form the basic development suite (or toolchain) which is used to build the actual system in Chapter 6. Some of these packages are needed to resolve circular dependencies—for example, to compile a compiler, you need a compiler.

Chapter 5 also shows you how to build a first pass of the toolchain, including Binutils and GCC (first pass basically means these two core packages will be reinstalled). The next step is to build Glibc, the C library. Glibc will be compiled by the toolchain programs built in the first pass. Then, a second pass of the toolchain will be built. This time, the toolchain will be dynamically linked against the newly built Glibc. The remaining Chapter 5 packages are built using this second pass toolchain. When this is done, the LFS installation process will no longer depend on the host distribution, with the exception of the running kernel.

This effort to isolate the new system from the host distribution may seem excessive. A full technical explanation as to why this is done is provided in Section 5.2, “Toolchain Technical Notes”.

In Chapter 6, the full LFS system is built. The chroot (change root) program is used to enter a virtual environment and start a new shell whose root directory will be set to the LFS partition. This is very similar to rebooting and instructing the kernel to mount the LFS partition as the root partition. The system does not actually reboot, but instead uses chroot because creating a bootable system requires additional work which is not necessary just yet. The major advantage is that “chrooting” allows you to continue using the host system while LFS is being built. While waiting for package compilations to complete, you can continue using your computer as normal.

To finish the installation, the basic system configuration is set up in Chapter 7, and the kernel and boot loader are set up in Chapter 8. Chapter 9 contains information on continuing the LFS experience beyond this book. After the steps in this book have been implemented, the computer will be ready to reboot into the new LFS system.

This is the process in a nutshell. Detailed information on each step is discussed in the following chapters and package descriptions. Items that may seem complicated will be clarified, and everything will fall into place as you embark on the LFS adventure.

Below is a list of package updates made since the previous release of the book.

This is version 8.1-rc2-systemd of the Linux From Scratch book, dated August 18, 2017. If this book is more than six months old, a newer and better version is probably already available. To find out, please check one of the mirrors via http://www.linuxfromscratch.org/mirrors.html.

Below is a list of changes made since the previous release of the book.

If an issue or a question is encountered while working through this book, please check the FAQ page at http://www.linuxfromscratch.org/faq/#generalfaq. Questions are often already answered there. If your question is not answered on this page, try to find the source of the problem. The following hint will give you some guidance for troubleshooting: http://www.linuxfromscratch.org/hints/downloads/files/errors.txt.

If you cannot find your problem listed in the FAQ, search the mailing lists at http://www.linuxfromscratch.org/search.html.

We also have a wonderful LFS community that is willing to offer assistance through the mailing lists and IRC (see the Section 1.4, “Resources” section of this book). However, we get several support questions every day and many of them can be easily answered by going to the FAQ and by searching the mailing lists first. So, for us to offer the best assistance possible, you need to do some research on your own first. That allows us to focus on the more unusual support needs. If your searches do not produce a solution, please include all relevant information (mentioned below) in your request for help.

gcc -DALIASPATH=\"/mnt/lfs/usr/share/locale:.\"
-DLOCALEDIR=\"/mnt/lfs/usr/share/locale\"
-DLIBDIR=\"/mnt/lfs/usr/lib\"
-DINCLUDEDIR=\"/mnt/lfs/usr/include\" -DHAVE_CONFIG_H -I. -I.
-g -O2 -c getopt1.c
gcc -g -O2 -static -o make ar.o arscan.o commands.o dir.o
expand.o file.o function.o getopt.o implicit.o job.o main.o
misc.o read.o remake.o rule.o signame.o variable.o vpath.o
default.o remote-stub.o version.o opt1.o
-lutil job.o: In function `load_too_high':
/lfs/tmp/make-3.79.1/job.c:1565: undefined reference
to `getloadavg'
collect2: ld returned 1 exit status
make[2]: *** [make] Error 1
make[2]: Leaving directory `/lfs/tmp/make-3.79.1'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/lfs/tmp/make-3.79.1'
make: *** [all-recursive-am] Error 2

make [2]: *** [make] Error 1

In this chapter, the host tools needed for building LFS are checked and, if necessary, installed. Then a partition which will host the LFS system is prepared. We will create the partition itself, create a file system on it, and mount it.

Your host system should have the following software with the minimum versions indicated. This should not be an issue for most modern Linux distributions. Also note that many distributions will place software headers into separate packages, often in the form of “<package-name>-devel” or “<package-name>-dev”. Be sure to install those if your distribution provides them.

Earlier versions of the listed software packages may work, but have not been tested.

To see whether your host system has all the appropriate versions, and the ability to compile programs, run the following:

cat > version-check.sh << "EOF"
#!/bin/bash
# Simple script to list version numbers of critical development tools
export LC_ALL=C
bash --version | head -n1 | cut -d" " -f2-4
MYSH=$(readlink -f /bin/sh)
echo "/bin/sh -> $MYSH"
echo $MYSH | grep -q bash || echo "ERROR: /bin/sh does not point to bash"
unset MYSH

echo -n "Binutils: "; ld --version | head -n1 | cut -d" " -f3-
bison --version | head -n1

if [ -h /usr/bin/yacc ]; then
  echo "/usr/bin/yacc -> `readlink -f /usr/bin/yacc`";
elif [ -x /usr/bin/yacc ]; then
  echo yacc is `/usr/bin/yacc --version | head -n1`
else
  echo "yacc not found" 
fi

bzip2 --version 2>&1 < /dev/null | head -n1 | cut -d" " -f1,6-
echo -n "Coreutils: "; chown --version | head -n1 | cut -d")" -f2
diff --version | head -n1
find --version | head -n1
gawk --version | head -n1

if [ -h /usr/bin/awk ]; then
  echo "/usr/bin/awk -> `readlink -f /usr/bin/awk`";
elif [ -x /usr/bin/awk ]; then
  echo awk is `/usr/bin/awk --version | head -n1`
else 
  echo "awk not found" 
fi

gcc --version | head -n1
g++ --version | head -n1
ldd --version | head -n1 | cut -d" " -f2-  # glibc version
grep --version | head -n1
gzip --version | head -n1
cat /proc/version
m4 --version | head -n1
make --version | head -n1
patch --version | head -n1
echo Perl `perl -V:version`
sed --version | head -n1
tar --version | head -n1
makeinfo --version | head -n1
xz --version | head -n1

echo 'int main(){}' > dummy.c && g++ -o dummy dummy.c
if [ -x dummy ]
  then echo "g++ compilation OK";
  else echo "g++ compilation failed"; fi
rm -f dummy.c dummy
EOF

bash version-check.sh

LFS is designed to be built in one session. That is, the instructions assume that the system will not be shut down during the process. That does not mean that the system has to be done in one sitting. The issue is that certain procedures have to be re-accomplished after a reboot if resuming LFS at different points.

Like most other operating systems, LFS is usually installed on a dedicated partition. The recommended approach to building an LFS system is to use an available empty partition or, if you have enough unpartitioned space, to create one.

A minimal system requires a partition of around 6 gigabytes (GB). This is enough to store all the source tarballs and compile the packages. However, if the LFS system is intended to be the primary Linux system, additional software will probably be installed which will require additional space. A 20 GB partition is a reasonable size to provide for growth. The LFS system itself will not take up this much room. A large portion of this requirement is to provide sufficient free temporary storage as well as for adding additional capabilities after LFS is complete. Additionally, compiling packages can require a lot of disk space which will be reclaimed after the package is installed.

Because there is not always enough Random Access Memory (RAM) available for compilation processes, it is a good idea to use a small disk partition as swap space. This is used by the kernel to store seldom-used data and leave more memory available for active processes. The swap partition for an LFS system can be the same as the one used by the host system, in which case it is not necessary to create another one.

Start a disk partitioning program such as cfdisk or fdisk with a command line option naming the hard disk on which the new partition will be created—for example /dev/sda for the primary Integrated Drive Electronics (IDE) disk. Create a Linux native partition and a swap partition, if needed. Please refer to cfdisk(8) or fdisk(8) if you do not yet know how to use the programs.

Remember the designation of the new partition (e.g., sda5). This book will refer to this as the LFS partition. Also remember the designation of the swap partition. These names will be needed later for the /etc/fstab file.

Now that a blank partition has been set up, the file system can be created. LFS can use any file system recognized by the Linux kernel, but the most common types are ext3 and ext4. The choice of file system can be complex and depends on the characteristics of the files and the size of the partition. For example:

Other file systems, including FAT32, NTFS, ReiserFS, JFS, and XFS are useful for specialized purposes. More information about these file systems can be found at http://en.wikipedia.org/wiki/Comparison_of_file_systems.

LFS assumes that the root file system (/) is of type ext4. To create an ext4 file system on the LFS partition, run the following:

mkfs -v -t ext4 /dev/<xxx>

If you are using an existing swap partition, there is no need to format it. If a new swap partition was created, it will need to be initialized with this command:

mkswap /dev/<yyy>

Replace <yyy> with the name of the swap partition.

Throughout this book, the environment variable LFS will be used several times. You should ensure that this variable is always defined throughout the LFS build process. It should be set to the name of the directory where you will be building your LFS system - we will use /mnt/lfs as an example, but the directory choice is up to you. If you are building LFS on a separate partition, this directory will be the mount point for the partition. Choose a directory location and set the variable with the following command:

export LFS=/mnt/lfs

Having this variable set is beneficial in that commands such as mkdir -v $LFS/tools can be typed literally. The shell will automatically replace “$LFS” with “/mnt/lfs” (or whatever the variable was set to) when it processes the command line.

echo $LFS

Now that a file system has been created, the partition needs to be made accessible. In order to do this, the partition needs to be mounted at a chosen mount point. For the purposes of this book, it is assumed that the file system is mounted under the directory specified by the LFS environment variable as described in the previous section.

Create the mount point and mount the LFS file system by running:

mkdir -pv $LFS
mount -v -t ext4 /dev/<xxx> $LFS

Replace <xxx> with the designation of the LFS partition.

If using multiple partitions for LFS (e.g., one for / and another for /usr), mount them using:

mkdir -pv $LFS
mount -v -t ext4 /dev/<xxx> $LFS
mkdir -v $LFS/usr
mount -v -t ext4 /dev/<yyy> $LFS/usr

Replace <xxx> and <yyy> with the appropriate partition names.

Ensure that this new partition is not mounted with permissions that are too restrictive (such as the nosuid or nodev options). Run the mount command without any parameters to see what options are set for the mounted LFS partition. If nosuid and/or nodev are set, the partition will need to be remounted.

If you are using a swap partition, ensure that it is enabled using the swapon command:

/sbin/swapon -v /dev/<zzz>

Replace <zzz> with the name of the swap partition.

Now that there is an established place to work, it is time to download the packages.

This chapter includes a list of packages that need to be downloaded in order to build a basic Linux system. The listed version numbers correspond to versions of the software that are known to work, and this book is based on their use. We highly recommend against using newer versions because the build commands for one version may not work with a newer version. The newest package versions may also have problems that require work-arounds. These work-arounds will be developed and stabilized in the development version of the book.

Download locations may not always be accessible. If a download location has changed since this book was published, Google (http://www.google.com/) provides a useful search engine for most packages. If this search is unsuccessful, try one of the alternative means of downloading discussed at http://www.linuxfromscratch.org/lfs/packages.html#packages.

Downloaded packages and patches will need to be stored somewhere that is conveniently available throughout the entire build. A working directory is also required to unpack the sources and build them. $LFS/sources can be used both as the place to store the tarballs and patches and as a working directory. By using this directory, the required elements will be located on the LFS partition and will be available during all stages of the building process.

To create this directory, execute the following command, as user root, before starting the download session:

mkdir -v $LFS/sources

Make this directory writable and sticky. “Sticky” means that even if multiple users have write permission on a directory, only the owner of a file can delete the file within a sticky directory. The following command will enable the write and sticky modes:

chmod -v a+wt $LFS/sources

An easy way to download all of the packages and patches is by using wget-list as an input to wget. For example:

wget --input-file=wget-list --continue --directory-prefix=$LFS/sources

Additionally, starting with LFS-7.0, there is a separate file, md5sums, which can be used to verify that all the correct packages are available before proceeding. Place that file in $LFS/sources and run:

pushd $LFS/sources
md5sum -c md5sums
popd

Download or otherwise obtain the following packages:

Total size of these packages: about 328 MB

In addition to the packages, several patches are also required. These patches correct any mistakes in the packages that should be fixed by the maintainer. The patches also make small modifications to make the packages easier to work with. The following patches will be needed to build an LFS system:

Total size of these patches: about 201.4 KB

In addition to the above required patches, there exist a number of optional patches created by the LFS community. These optional patches solve minor problems or enable functionality that is not enabled by default. Feel free to peruse the patches database located at http://www.linuxfromscratch.org/patches/downloads/ and acquire any additional patches to suit your system needs.

In this chapter, we will perform a few additional tasks to prepare for building the temporary system. We will create a directory in $LFS for the installation of the temporary tools, add an unprivileged user to reduce risk, and create an appropriate build environment for that user. We will also explain the unit of time we use to measure how long LFS packages take to build, or “SBUs”, and give some information about package test suites.

All programs compiled in Chapter 5 will be installed under $LFS/tools to keep them separate from the programs compiled in Chapter 6. The programs compiled here are temporary tools and will not be a part of the final LFS system. By keeping these programs in a separate directory, they can easily be discarded later after their use. This also prevents these programs from ending up in the host production directories (easy to do by accident in Chapter 5).

Create the required directory by running the following as root:

mkdir -v $LFS/tools

The next step is to create a /tools symlink on the host system. This will point to the newly-created directory on the LFS partition. Run this command as root as well:

ln -sv $LFS/tools /

The created symlink enables the toolchain to be compiled so that it always refers to /tools, meaning that the compiler, assembler, and linker will work both in Chapter 5 (when we are still using some tools from the host) and in the next (when we are “chrooted” to the LFS partition).

When logged in as user root, making a single mistake can damage or destroy a system. Therefore, we recommend building the packages in this chapter as an unprivileged user. You could use your own user name, but to make it easier to set up a clean working environment, create a new user called lfs as a member of a new group (also named lfs) and use this user during the installation process. As root, issue the following commands to add the new user:

groupadd lfs
useradd -s /bin/bash -g lfs -m -k /dev/null lfs

To log in as lfs (as opposed to switching to user lfs when logged in as root, which does not require the lfs user to have a password), give lfs a password:

passwd lfs

Grant lfs full access to $LFS/tools by making lfs the directory owner:

chown -v lfs $LFS/tools

If a separate working directory was created as suggested, give user lfs ownership of this directory:

chown -v lfs $LFS/sources

Next, login as user lfs. This can be done via a virtual console, through a display manager, or with the following substitute user command:

su - lfs

The “-” instructs su to start a login shell as opposed to a non-login shell. The difference between these two types of shells can be found in detail in bash(1) and info bash.

Set up a good working environment by creating two new startup files for the bash shell. While logged in as user lfs, issue the following command to create a new .bash_profile:

cat > ~/.bash_profile << "EOF"
exec env -i HOME=$HOME TERM=$TERM PS1='\u:\w\$ ' /bin/bash
EOF

When logged on as user lfs, the initial shell is usually a login shell which reads the /etc/profile of the host (probably containing some settings and environment variables) and then .bash_profile. The exec env -i.../bin/bash command in the .bash_profile file replaces the running shell with a new one with a completely empty environment, except for the HOME, TERM, and PS1 variables. This ensures that no unwanted and potentially hazardous environment variables from the host system leak into the build environment. The technique used here achieves the goal of ensuring a clean environment.

The new instance of the shell is a non-login shell, which does not read the /etc/profile or .bash_profile files, but rather reads the .bashrc file instead. Create the .bashrc file now:

cat > ~/.bashrc << "EOF"
set +h
umask 022
LFS=/mnt/lfs
LC_ALL=POSIX
LFS_TGT=$(uname -m)-lfs-linux-gnu
PATH=/tools/bin:/bin:/usr/bin
export LFS LC_ALL LFS_TGT PATH
EOF

The set +h command turns off bash's hash function. Hashing is ordinarily a useful feature—bash uses a hash table to remember the full path of executable files to avoid searching the PATH time and again to find the same executable. However, the new tools should be used as soon as they are installed. By switching off the hash function, the shell will always search the PATH when a program is to be run. As such, the shell will find the newly compiled tools in $LFS/tools as soon as they are available without remembering a previous version of the same program in a different location.

Setting the user file-creation mask (umask) to 022 ensures that newly created files and directories are only writable by their owner, but are readable and executable by anyone (assuming default modes are used by the open(2) system call, new files will end up with permission mode 644 and directories with mode 755).

The LFS variable should be set to the chosen mount point.

The LC_ALL variable controls the localization of certain programs, making their messages follow the conventions of a specified country. Setting LC_ALL to “POSIX” or “C” (the two are equivalent) ensures that everything will work as expected in the chroot environment.

The LFS_TGT variable sets a non-default, but compatible machine description for use when building our cross compiler and linker and when cross compiling our temporary toolchain. More information is contained in Section 5.2, “Toolchain Technical Notes”.

By putting /tools/bin ahead of the standard PATH, all the programs installed in Chapter 5 are picked up by the shell immediately after their installation. This, combined with turning off hashing, limits the risk that old programs are used from the host when the same programs are available in the chapter 5 environment.

Finally, to have the environment fully prepared for building the temporary tools, source the just-created user profile:

source ~/.bash_profile

Many people would like to know beforehand approximately how long it takes to compile and install each package. Because Linux From Scratch can be built on many different systems, it is impossible to provide accurate time estimates. The biggest package (Glibc) will take approximately 20 minutes on the fastest systems, but could take up to three days on slower systems! Instead of providing actual times, the Standard Build Unit (SBU) measure will be used instead.

The SBU measure works as follows. The first package to be compiled from this book is Binutils in Chapter 5. The time it takes to compile this package is what will be referred to as the Standard Build Unit or SBU. All other compile times will be expressed relative to this time.

For example, consider a package whose compilation time is 4.5 SBUs. This means that if a system took 10 minutes to compile and install the first pass of Binutils, it will take approximately 45 minutes to build this example package. Fortunately, most build times are shorter than the one for Binutils.

In general, SBUs are not entirely accurate because they depend on many factors, including the host system's version of GCC. They are provided here to give an estimate of how long it might take to install a package, but the numbers can vary by as much as dozens of minutes in some cases.

export MAKEFLAGS='-j 2'

make -j2

Most packages provide a test suite. Running the test suite for a newly built package is a good idea because it can provide a “sanity check” indicating that everything compiled correctly. A test suite that passes its set of checks usually proves that the package is functioning as the developer intended. It does not, however, guarantee that the package is totally bug free.

Some test suites are more important than others. For example, the test suites for the core toolchain packages—GCC, Binutils, and Glibc—are of the utmost importance due to their central role in a properly functioning system. The test suites for GCC and Glibc can take a very long time to complete, especially on slower hardware, but are strongly recommended.

A common issue with running the test suites for Binutils and GCC is running out of pseudo terminals (PTYs). This can result in a high number of failing tests. This may happen for several reasons, but the most likely cause is that the host system does not have the devpts file system set up correctly. This issue is discussed in greater detail at http://www.linuxfromscratch.org/lfs/faq.html#no-ptys.

Sometimes package test suites will fail, but for reasons which the developers are aware of and have deemed non-critical. Consult the logs located at http://www.linuxfromscratch.org/lfs/build-logs/8.1-rc2/ to verify whether or not these failures are expected. This site is valid for all tests throughout this book.

This chapter shows how to build a minimal Linux system. This system will contain just enough tools to start constructing the final LFS system in Chapter 6 and allow a working environment with more user convenience than a minimum environment would.

There are two steps in building this minimal system. The first step is to build a new and host-independent toolchain (compiler, assembler, linker, libraries, and a few useful utilities). The second step uses this toolchain to build the other essential tools.

The files compiled in this chapter will be installed under the $LFS/tools directory to keep them separate from the files installed in the next chapter and the host production directories. Since the packages compiled here are temporary, we do not want them to pollute the soon-to-be LFS system.

This section explains some of the rationale and technical details behind the overall build method. It is not essential to immediately understand everything in this section. Most of this information will be clearer after performing an actual build. This section can be referred to at any time during the process.

The overall goal of Chapter 5 is to produce a temporary area that contains a known-good set of tools that can be isolated from the host system. By using chroot, the commands in the remaining chapters will be contained within that environment, ensuring a clean, trouble-free build of the target LFS system. The build process has been designed to minimize the risks for new readers and to provide the most educational value at the same time.

Some key technical points of how the Chapter 5 build method works:

Binutils is installed first because the configure runs of both GCC and Glibc perform various feature tests on the assembler and linker to determine which software features to enable or disable. This is more important than one might first realize. An incorrectly configured GCC or Glibc can result in a subtly broken toolchain, where the impact of such breakage might not show up until near the end of the build of an entire distribution. A test suite failure will usually highlight this error before too much additional work is performed.

Binutils installs its assembler and linker in two locations, /tools/bin and /tools/$LFS_TGT/bin. The tools in one location are hard linked to the other. An important facet of the linker is its library search order. Detailed information can be obtained from ld by passing it the --verbose flag. For example, an ld --verbose | grep SEARCH will illustrate the current search paths and their order. It shows which files are linked by ld by compiling a dummy program and passing the --verbose switch to the linker. For example, gcc dummy.c -Wl,--verbose 2>&1 | grep succeeded will show all the files successfully opened during the linking.

The next package installed is GCC. An example of what can be seen during its run of configure is:

checking what assembler to use... /tools/i686-lfs-linux-gnu/bin/as
checking what linker to use... /tools/i686-lfs-linux-gnu/bin/ld

This is important for the reasons mentioned above. It also demonstrates that GCC's configure script does not search the PATH directories to find which tools to use. However, during the actual operation of gcc itself, the same search paths are not necessarily used. To find out which standard linker gcc will use, run: gcc -print-prog-name=ld.

Detailed information can be obtained from gcc by passing it the -v command line option while compiling a dummy program. For example, gcc -v dummy.c will show detailed information about the preprocessor, compilation, and assembly stages, including gcc's included search paths and their order.

Next installed are sanitized Linux API headers. These allow the standard C library (Glibc) to interface with features that the Linux kernel will provide.

The next package installed is Glibc. The most important considerations for building Glibc are the compiler, binary tools, and kernel headers. The compiler is generally not an issue since Glibc will always use the compiler relating to the --host parameter passed to its configure script; e.g. in our case, the compiler will be i686-lfs-linux-gnu-gcc. The binary tools and kernel headers can be a bit more complicated. Therefore, take no risks and use the available configure switches to enforce the correct selections. After the run of configure, check the contents of the config.make file in the glibc-build directory for all important details. Note the use of CC="i686-lfs-gnu-gcc" to control which binary tools are used and the use of the -nostdinc and -isystem flags to control the compiler's include search path. These items highlight an important aspect of the Glibc package—it is very self-sufficient in terms of its build machinery and generally does not rely on toolchain defaults.

During the second pass of Binutils, we are able to utilize the --with-lib-path configure switch to control ld's library search path.

For the second pass of GCC, its sources also need to be modified to tell GCC to use the new dynamic linker. Failure to do so will result in the GCC programs themselves having the name of the dynamic linker from the host system's /lib directory embedded into them, which would defeat the goal of getting away from the host. From this point onwards, the core toolchain is self-contained and self-hosted. The remainder of the Chapter 5 packages all build against the new Glibc in /tools.

Upon entering the chroot environment in Chapter 6, the first major package to be installed is Glibc, due to its self-sufficient nature mentioned above. Once this Glibc is installed into /usr, we will perform a quick changeover of the toolchain defaults, and then proceed in building the rest of the target LFS system.

When building packages there are several assumptions made within the instructions:

The steps in this section are optional, but if the LFS partition is rather small, it is beneficial to learn that unnecessary items can be removed. The executables and libraries built so far contain about 70 MB of unneeded debugging symbols. Remove those symbols with:

strip --strip-debug /tools/lib/*
/usr/bin/strip --strip-unneeded /tools/{,s}bin/*

These commands will skip a number of files, reporting that it does not recognize their file format. Most of these are scripts instead of binaries. Also use the system strip command to include the strip binary in /tools.

Take care not to use --strip-unneeded on the libraries. The static ones would be destroyed and the toolchain packages would need to be built all over again.

To save more, remove the documentation:

rm -rf /tools/{,share}/{info,man,doc}

At this point, you should have at least 3 GB of free space in $LFS that can be used to build and install Glibc and Gcc in the next phase. If you can build and install Glibc, you can build and install the rest too.

Currently, the $LFS/tools directory is owned by the user lfs, a user that exists only on the host system. If the $LFS/tools directory is kept as is, the files are owned by a user ID without a corresponding account. This is dangerous because a user account created later could get this same user ID and would own the $LFS/tools directory and all the files therein, thus exposing these files to possible malicious manipulation.

To avoid this issue, you could add the lfs user to the new LFS system later when creating the /etc/passwd file, taking care to assign it the same user and group IDs as on the host system. Better yet, change the ownership of the $LFS/tools directory to user root by running the following command:

chown -R root:root $LFS/tools

Although the $LFS/tools directory can be deleted once the LFS system has been finished, it can be retained to build additional LFS systems of the same book version. How best to backup $LFS/tools is a matter of personal preference.

In this chapter, we enter the building site and start constructing the LFS system in earnest. That is, we chroot into the temporary mini Linux system, make a few final preparations, and then begin installing the packages.

The installation of this software is straightforward. Although in many cases the installation instructions could be made shorter and more generic, we have opted to provide the full instructions for every package to minimize the possibilities for mistakes. The key to learning what makes a Linux system work is to know what each package is used for and why you (or the system) may need it.

We do not recommend using optimizations. They can make a program run slightly faster, but they may also cause compilation difficulties and problems when running the program. If a package refuses to compile when using optimization, try to compile it without optimization and see if that fixes the problem. Even if the package does compile when using optimization, there is the risk it may have been compiled incorrectly because of the complex interactions between the code and build tools. Also note that the -march and -mtune options using values not specified in the book have not been tested. This may cause problems with the toolchain packages (Binutils, GCC and Glibc). The small potential gains achieved in using compiler optimizations are often outweighed by the risks. First-time builders of LFS are encouraged to build without custom optimizations. The subsequent system will still run very fast and be stable at the same time.

The order that packages are installed in this chapter needs to be strictly followed to ensure that no program accidentally acquires a path referring to /tools hard-wired into it. For the same reason, do not compile separate packages in parallel. Compiling in parallel may save time (especially on dual-CPU machines), but it could result in a program containing a hard-wired path to /tools, which will cause the program to stop working when that directory is removed.

Before the installation instructions, each installation page provides information about the package, including a concise description of what it contains, approximately how long it will take to build, and how much disk space is required during this building process. Following the installation instructions, there is a list of programs and libraries (along with brief descriptions of these) that the package installs.

Various file systems exported by the kernel are used to communicate to and from the kernel itself. These file systems are virtual in that no disk space is used for them. The content of the file systems resides in memory.

Begin by creating directories onto which the file systems will be mounted:

mkdir -pv $LFS/{dev,proc,sys,run}

mknod -m 600 $LFS/dev/console c 5 1
mknod -m 666 $LFS/dev/null c 1 3

mount -v --bind /dev $LFS/dev

mount -vt devpts devpts $LFS/dev/pts -o gid=5,mode=620
mount -vt proc proc $LFS/proc
mount -vt sysfs sysfs $LFS/sys
mount -vt tmpfs tmpfs $LFS/run

if [ -h $LFS/dev/shm ]; then
  mkdir -pv $LFS/$(readlink $LFS/dev/shm)
fi

Package Management is an often requested addition to the LFS Book. A Package Manager allows tracking the installation of files making it easy to remove and upgrade packages. As well as the binary and library files, a package manager will handle the installation of configuration files. Before you begin to wonder, NO—this section will not talk about nor recommend any particular package manager. What it provides is a roundup of the more popular techniques and how they work. The perfect package manager for you may be among these techniques or may be a combination of two or more of these techniques. This section briefly mentions issues that may arise when upgrading packages.

Some reasons why no package manager is mentioned in LFS or BLFS include:

There are some hints written on the topic of package management. Visit the Hints Project and see if one of them fits your need.

./configure --prefix=/usr/pkg/libfoo/1.1
make
make install

./configure --prefix=/usr
make
make DESTDIR=/usr/pkg/libfoo/1.1 install

It is time to enter the chroot environment to begin building and installing the final LFS system. As user root, run the following command to enter the realm that is, at the moment, populated with only the temporary tools:

chroot "$LFS" /tools/bin/env -i \
    HOME=/root                  \
    TERM="$TERM"                \
    PS1='\u:\w\$ '              \
    PATH=/bin:/usr/bin:/sbin:/usr/sbin:/tools/bin \
    /tools/bin/bash --login +h

The -i option given to the env command will clear all variables of the chroot environment. After that, only the HOME, TERM, PS1, and PATH variables are set again. The TERM=$TERM construct will set the TERM variable inside chroot to the same value as outside chroot. This variable is needed for programs like vim and less to operate properly. If other variables are needed, such as CFLAGS or CXXFLAGS, this is a good place to set them again.

From this point on, there is no need to use the LFS variable anymore, because all work will be restricted to the LFS file system. This is because the Bash shell is told that $LFS is now the root (/) directory.

Notice that /tools/bin comes last in the PATH. This means that a temporary tool will no longer be used once its final version is installed. This occurs when the shell does not “remember” the locations of executed binaries—for this reason, hashing is switched off by passing the +h option to bash.

Note that the bash prompt will say I have no name! This is normal because the /etc/passwd file has not been created yet.

It is time to create some structure in the LFS file system. Create a standard directory tree by issuing the following commands:

mkdir -pv /{bin,boot,etc/{opt,sysconfig},home,lib/firmware,mnt,opt}
mkdir -pv /{media/{floppy,cdrom},sbin,srv,var}
install -dv -m 0750 /root
install -dv -m 1777 /tmp /var/tmp
mkdir -pv /usr/{,local/}{bin,include,lib,sbin,src}
mkdir -pv /usr/{,local/}share/{color,dict,doc,info,locale,man}
mkdir -v  /usr/{,local/}share/{misc,terminfo,zoneinfo}
mkdir -v  /usr/libexec
mkdir -pv /usr/{,local/}share/man/man{1..8}

case $(uname -m) in
 x86_64) mkdir -v /lib64 ;;
esac

mkdir -v /var/{log,mail,spool}
ln -sv /run /var/run
ln -sv /run/lock /var/lock
mkdir -pv /var/{opt,cache,lib/{color,misc,locate},local}

Directories are, by default, created with permission mode 755, but this is not desirable for all directories. In the commands above, two changes are made—one to the home directory of user root, and another to the directories for temporary files.

The first mode change ensures that not just anybody can enter the /root directory—the same as a normal user would do with his or her home directory. The second mode change makes sure that any user can write to the /tmp and /var/tmp directories, but cannot remove another user's files from them. The latter is prohibited by the so-called “sticky bit,” the highest bit (1) in the 1777 bit mask.

Some programs use hard-wired paths to programs which do not exist yet. In order to satisfy these programs, create a number of symbolic links which will be replaced by real files throughout the course of this chapter after the software has been installed:

ln -sv /tools/bin/{bash,cat,dd,echo,ln,pwd,rm,stty} /bin
ln -sv /tools/bin/{install,perl} /usr/bin
ln -sv /tools/lib/libgcc_s.so{,.1} /usr/lib
ln -sv /tools/lib/libstdc++.{a,so{,.6}} /usr/lib
sed 's/tools/usr/' /tools/lib/libstdc++.la > /usr/lib/libstdc++.la
for lib in blkid lzma mount uuid
do
    ln -sv /tools/lib/lib$lib.{a,so*} /usr/lib
    sed 's/tools/usr/' /tools/lib/lib${lib}.la > /usr/lib/lib${lib}.la
done
ln -sv bash /bin/sh

Historically, Linux maintains a list of the mounted file systems in the file /etc/mtab. Modern kernels maintain this list internally and exposes it to the user via the /proc filesystem. To satisfy utilities that expect the presence of /etc/mtab, create the following symbolic link:

ln -sv /proc/self/mounts /etc/mtab

In order for user root to be able to login and for the name “root” to be recognized, there must be relevant entries in the /etc/passwd and /etc/group files.

Create the /etc/passwd file by running the following command:

cat > /etc/passwd << "EOF"
root:x:0:0:root:/root:/bin/bash
bin:x:1:1:bin:/dev/null:/bin/false
daemon:x:6:6:Daemon User:/dev/null:/bin/false
messagebus:x:18:18:D-Bus Message Daemon User:/var/run/dbus:/bin/false
systemd-bus-proxy:x:72:72:systemd Bus Proxy:/:/bin/false
systemd-journal-gateway:x:73:73:systemd Journal Gateway:/:/bin/false
systemd-journal-remote:x:74:74:systemd Journal Remote:/:/bin/false
systemd-journal-upload:x:75:75:systemd Journal Upload:/:/bin/false
systemd-network:x:76:76:systemd Network Management:/:/bin/false
systemd-resolve:x:77:77:systemd Resolver:/:/bin/false
systemd-timesync:x:78:78:systemd Time Synchronization:/:/bin/false
systemd-coredump:x:79:79:systemd Core Dumper:/:/bin/false
nobody:x:99:99:Unprivileged User:/dev/null:/bin/false
EOF

The actual password for root (the “x” used here is just a placeholder) will be set later.

Create the /etc/group file by running the following command:

cat > /etc/group << "EOF"
root:x:0:
bin:x:1:daemon
sys:x:2:
kmem:x:3:
tape:x:4:
tty:x:5:
daemon:x:6:
floppy:x:7:
disk:x:8:
lp:x:9:
dialout:x:10:
audio:x:11:
video:x:12:
utmp:x:13:
usb:x:14:
cdrom:x:15:
adm:x:16:
messagebus:x:18:
systemd-journal:x:23:
input:x:24:
mail:x:34:
systemd-bus-proxy:x:72:
systemd-journal-gateway:x:73:
systemd-journal-remote:x:74:
systemd-journal-upload:x:75:
systemd-network:x:76:
systemd-resolve:x:77:
systemd-timesync:x:78:
systemd-coredump:x:79:
nogroup:x:99:
users:x:999:
EOF

The created groups are not part of any standard—they are groups decided on in part by the requirements of the Udev configuration in this chapter, and in part by common convention employed by a number of existing Linux distributions. In addition, some test suites rely on specific users or groups. The Linux Standard Base (LSB, available at http://www.linuxbase.org) recommends only that, besides the group root with a Group ID (GID) of 0, a group bin with a GID of 1 be present. All other group names and GIDs can be chosen freely by the system administrator since well-written programs do not depend on GID numbers, but rather use the group's name.

To remove the “I have no name!” prompt, start a new shell. Since a full Glibc was installed in Chapter 5 and the /etc/passwd and /etc/group files have been created, user name and group name resolution will now work:

exec /tools/bin/bash --login +h

Note the use of the +h directive. This tells bash not to use its internal path hashing. Without this directive, bash would remember the paths to binaries it has executed. To ensure the use of the newly compiled binaries as soon as they are installed, the +h directive will be used for the duration of this chapter.

The login, agetty, and init programs (and others) use a number of log files to record information such as who was logged into the system and when. However, these programs will not write to the log files if they do not already exist. Initialize the log files and give them proper permissions:

touch /var/log/{btmp,lastlog,faillog,wtmp}
chgrp -v utmp /var/log/lastlog
chmod -v 664  /var/log/lastlog
chmod -v 600  /var/log/btmp

The /var/log/wtmp file records all logins and logouts. The /var/log/lastlog file records when each user last logged in. The /var/log/faillog file records failed login attempts. The /var/log/btmp file records the bad login attempts.

Now that the final C libraries have been installed, it is time to adjust the toolchain so that it will link any newly compiled program against these new libraries.

First, backup the /tools linker, and replace it with the adjusted linker we made in chapter 5. We'll also create a link to its counterpart in /tools/$(uname -m)-pc-linux-gnu/bin:

mv -v /tools/bin/{ld,ld-old}
mv -v /tools/$(uname -m)-pc-linux-gnu/bin/{ld,ld-old}
mv -v /tools/bin/{ld-new,ld}
ln -sv /tools/bin/ld /tools/$(uname -m)-pc-linux-gnu/bin/ld

Next, amend the GCC specs file so that it points to the new dynamic linker. Simply deleting all instances of “/tools” should leave us with the correct path to the dynamic linker. Also adjust the specs file so that GCC knows where to find the correct headers and Glibc start files. A sed command accomplishes this:

gcc -dumpspecs | sed -e 's@/tools@@g'                   \
    -e '/\*startfile_prefix_spec:/{n;s@.*@/usr/lib/ @}' \
    -e '/\*cpp:/{n;s@$@ -isystem /usr/include@}' >      \
    `dirname $(gcc --print-libgcc-file-name)`/specs

It is a good idea to visually inspect the specs file to verify the intended change was actually made.

It is imperative at this point to ensure that the basic functions (compiling and linking) of the adjusted toolchain are working as expected. To do this, perform the following sanity checks:

echo 'int main(){}' > dummy.c
cc dummy.c -v -Wl,--verbose &> dummy.log
readelf -l a.out | grep ': /lib'

There should be no errors, and the output of the last command will be (allowing for platform-specific differences in dynamic linker name):

[Requesting program interpreter: /lib64/ld-linux-x86-64.so.2]

Note that on 64-bit systems /lib is the location of our dynamic linker, but is accessed via a symbolic link in /lib64.

Now make sure that we're setup to use the correct start files:

grep -o '/usr/lib.*/crt[1in].*succeeded' dummy.log

The output of the last command should be:

/usr/lib/../lib/crt1.o succeeded
/usr/lib/../lib/crti.o succeeded
/usr/lib/../lib/crtn.o succeeded

Verify that the compiler is searching for the correct header files:

grep -B1 '^ /usr/include' dummy.log

This command should return the following output:

#include <...> search starts here:
 /usr/include

Next, verify that the new linker is being used with the correct search paths:

grep 'SEARCH.*/usr/lib' dummy.log |sed 's|; |\n|g'

References to paths that have components with '-linux-gnu' should be ignored, but otherwise the output of the last command should be:

SEARCH_DIR("/usr/lib")
SEARCH_DIR("/lib")

Next make sure that we're using the correct libc:

grep "/lib.*/libc.so.6 " dummy.log

The output of the last command should be:

attempt to open /lib/libc.so.6 succeeded

Lastly, make sure GCC is using the correct dynamic linker:

grep found dummy.log

The output of the last command should be (allowing for platform-specific differences in dynamic linker name):

found ld-linux-x86-64.so.2 at /lib/ld-linux-x86-64.so.2

If the output does not appear as shown above or is not received at all, then something is seriously wrong. Investigate and retrace the steps to find out where the problem is and correct it. The most likely reason is that something went wrong with the specs file adjustment. Any issues will need to be resolved before continuing on with the process.

Once everything is working correctly, clean up the test files:

rm -v dummy.c a.out dummy.log