Loadable Kernel Modules

J a n u a r y / F e b r u a r y 1 9 9 9 I E E E S o f t w a r e 6 5 ost Unix kernels are monolithic1,2; that is, the kernel is a (normally big) piece of compact code, in which all functions share a common space and are tightly related. When they need to be updated, they must be relinked and reinstalled and the system rebooted before the changes can take effect. This makes modifying them, by adding and testing new drivers, very difficult. The Linux kernel particularly suffered from this problem because of its wide distribution and cooperative development: it was difficult to isolate, test, and integrate the continual stream of changes, enhancements, and additions by people from around the world. Kernel developers soon realized that something had to be done to isolate and track modifications and to avoid frequent kernel builds and reboots. The community adopted its solution from the microkernel approach for writing operating systems, where many kernel functions are separate user-space components and communicate via microkernel facilities. Although Linux is not a microkernel, it does use loadable modules to simplify and shorten development time, make dynamic configuration easier, and save kernel memory.3 This gives a level of flexibility and efficiency not present in many other Unixes. Juan-Mariano de Goyeneche and Elena Apolinario Fernández de Sousa, Technical University of Madrid Loadable Kernel Modules The L inux kerne l uses dynamica l ly loadable mo dules to s impl i fy and shor ten deve lopment t ime, to make conf igurat ion eas ie r, and to save kerne l memor y. Th is g ives a leve l o f f lex ib i l i t y and ef f i c ienc y not present in many other Unixes. H ere’s how i t ’s done.