Normally, I wouldn't have the slightest idea about any of this, but we just covered this yesterday in my network security class. Here's some fun facts
Quote:
The monolithic approach defines a high-level virtual interface over the hardware, with a set of primitives or system calls to implement operating system services such as process management, concurrency, and memory management in several modules that run in supervisor mode.
Even if every module servicing these operations is separate from the whole, the code integration is very tight and difficult to do correctly, and, since all the modules run in the same address space, a bug in one module can bring down the whole system. However, when the implementation is complete and trustworthy, the tight internal integration of components allows the low-level features of the underlying system to be effectively utilized, making a good monolithic kernel highly efficient. Proponents of the monolithic kernel approach make the case that if code is incorrect, it does not belong in a kernel, and if it is, there is little advantage in the microkernel approach.
|
Monolithic kernels are the Linux and BSD distributions, Microkernels include things like Minix and AmigaOS, and there are hybrid kernels like BeOS and Windows 2000/XP/2003/Vista or Mac OS X.
The problem with monolithic kernels is this: let's say you go buy a new piece of hardware, plug it in, and install the driver that came on the CD with it. Now that driver is sitting in your kernel memory and has access to all kernel functions, all system memory, etc. Do you know who wrote the driver? Do you trust them with your data? Do you trust them to write a
bug free driver to not crash your system? Bad drivers are the fastest way to bring down a monolithic kernel, and there are a variety of other security flaws that can occur with them.
Microkernels on the other hand have as little as possible in kernel space: just the core OS functionality to handle interprocess communication, virtual memory, and thread management. Everything else, including drivers, run as userland processes. The problem with microkernels is that they're traditionally much slower than monolithic kernels, simply because it has to go to user memory every time it wants to write to a device, send a packet on the internet, etc.
Hybrid kernels are basically modified microkernels. They take the microkernel approach of having next to nothing in the kernel, but then adds some other code so that it performs better. For instance, they may decide to load the device drivers into system memory while still keeping the microkernel approach of segmenting everything else out of kernel memory.
So what's the better way of doing things? I honestly don't know
Either way, it's another fun fact about the "which OS is more/less secure" approach.