Why systems crash

NoCompanyIT · Aug 5, 2009

Oh man! Ad-blocking software has been detected! :'(

This website is run by the community, for the community... and it needs advertisements in order to keep running. Blocking our ads means your killing our stats!
Please disable your ad-block, or become a premium member to hide all advertisements and this notice.

I've just been reading loads of articles at serverwatch.com and came across an interesting one, which, as is the case with the internet, led me onto other sites, deeper and further.

I found this:

http://www.minix3.org/doc/reliable-os.pdf

A Lightweight Method for Building Reliable Operating Systems
Despite Unreliable Device Drivers
Technical Report IR-CS-018, January 2006

1.1 Why do Systems Crash?
The underlying reason that operating systems crash can be
traced back to two fundamental design flaws they all share:
too many privileges and lack of adequate fault isolation. Vir-
tually all operating systems consist of many modules linked
together in one address space to form a single binary pro-
gram that runs in kernel mode. A bug in any module can
easily trash critical data structures in an unrelated module
and bring the system down in an instant. The reason for
linking all the modules together in a single address space,
with no protection between the modules, is that designers
have made a Faustian bargain: better performance at the
cost of more system crashes. We will quantitatively examine
the price of this trade-off below

A closely related issue is why crashes occur in the first
place. After all, if every module were perfect, there would
be no need for fault isolation between the modules because
there would be no faults to propagate. We assert that most
faults are due to programming errors (bugs) and are largely
the consequence of too much complexity and the use of for-
eign code. Studies have shown that software averages 1-16
bugs per 1000 lines of code [27, 22, 2] and this range is
surely an underestimate because only bugs that were ulti-
mately found were counted. The obvious conclusion of these
studies is: more code means more bugs. As software devel-
ops, each new release tends to acquire more features (and
thus more code) and is often less reliable than its predeces-
sor. Studies [22] have shown that the number of bugs per
thousand lines of code tends to stabilize after many releases
and does not go asymptotically to zero.

Some of these bugs are exploited by attackers to allow
viruses and worms to infect and damage systems. Thus
some alleged security problems have nothing to do with the
security measures in principle (e.g., flawed cryptography or
broken authorization protocols), but are simply due to code
bugs (e.g., buffer overruns that allow execution of injected
code). In this paper, when we talk about reliability, we also
mean what is often referred to as securityunauthorized
access as a result of a bug in the code.

The second problem is introducing foreign code into the
operating system. Most sophisticated users would never al-
low a third party to insert unknown code into the heart of
their operating system, yet when they buy a new peripheral
device and install the driver, this is precisely what they are
doing. Device drivers are usually written by programmers
working for the peripheral manufacturer, which typically has
less quality control than the operating system vendor. When
the driver is an open-source effort, it is often authored by a
well-meaning but not necessarily experienced volunteer and
with even less quality control. In Linux, for example, the er-
ror rate on device drivers is three to seven times higher than
in the rest of the kernel [7]. Even Microsoft, which has the
motivation and resources to apply tighter quality controls,
does not do any better: 85% of all Windows XP crashes are
due to code bugs in device drivers [25].
Click to expand...

Log in or Sign up

Why systems crash

NoCompanyIT Nibble Poster

Share This Page

Navigation

Popular Forums

Useful Links

Log in or Sign up

Why systems crash

NoCompanyIT Nibble Poster

Share This Page

Useful Searches