Section 2.
Architecture.
"..of
course, Linus didn’t sit down in a vacuum and suddenly
type in the Linux source code.... He had my book.... But the
code was his. The proof of this is that he messed the design
up."
Andrew Tanenbaum.
2. Linux architecture.
The Linux kernel includes true
multitasking, virtual memory, shared libraries, shared
copy on write executables, memory management and
TCP/IP networking.
2.1. The kernel.
When I first began studying UNIX
the proud boast was that the kernel consisted of around
10,000 lines of C code and the hardware code represented
around 1000 lines of assembler. It was this compact size and
the use of a high level language that gave UNIX its
flexibility. By making the necessary changes to the hardware
code in assembler, UNIX could be relatively easily
recompiled for very different hardware platforms.
Linux has, in common with many
other operating systems and software applications, undergone
substantial code bloat.
In 2011 the Linux Foundation
kernel development study put the number of files used in
building the kernel at 37,000, with 17 million lines of
code.
Fortunately, although the kernel
remains monolithic, unlike many other operating systems, you
don’t need to keep it all. Many of the kernel virtual
devices are in the form of loadable modules, which can
either be loaded at boot time or later, on demand. In
addition, it is a relatively simple task, to rebuild the
kernel using only those static modules that are actually
needed in your system. The kernel can be slimmed down
substantially with significant performance gains.
The kernel runs in a highly
privileged mode and supervises the privileges of all other
processes.
Prior to kernel 2.2 there was a
simple bifurcation between processes that ran with the
effective user id 0 and which were then privileged and those
processes which ran with a non zero id and were therefore
unprivileged.
Starting with kernel 2.2 Linux
divides privileges formerly associated with the superusers
into units known as capabilities which can be enabled
and disabled independently.
It is the flexibility of the
kernel that allows it to be modified for such a large range
of devices.
Figure 1 - UNIX/Linux
Architecture
Figure 2 - System
Kernel
2.2. User programs.
Most users’ contact with
what they think of as the OS, is confined to the the huge
raft of software tools that come as standard with most
distributions of UNIX and Linux.
These software tools (and other
applications) communicate with the kernel through the
system call interface. Many of them were written
before Linus wrote the Linux kernel and were intended to be
part of the GNU Hurd operating system.
2.3. Exercises.
Andrew Tanenbaum believes that
Linus made fundamental errors in his design of Linux. Do
some on-line research and discover the key arguments over
the differences between Minix and Linux architecture.
Study the commands in the table
of tools below. Experiment with using each of them on the
command line and read the associated man pages.
2.4. Tools:
