Memory subsystem

If one compiles a curent Linux kernel one might easaly think it is not well suited for embedded systems - the linux kernel size has grown substantially between 2.0.X and 2.2.X and again between 2.2.X and 2.4.x. This has moved the minimum memmory demands up to 4MB weras a 2.0.X kernel could confortably operate with 2MB RAM. So is a 2.4.X based system not usable for embedded linux ? Not only did a rich and interesting set of featurs get added in the 2.4.X kernel series, notably the clean integration of MTD (Memory TEchnology Devices) and iproute2/QOS but also the way the kernel manages memmory resources has improved substantially and that is why even for resource constraint systems a 2.4.X kernel will performe better than a slim 2.0.X kernel. Major improvements are in the buffering mechanism the cleanup of cache-allignment and direct access to peripheral buffers from userspace (kobuf) and other low level extensions. It's not posible to describe the full memmory subsystem in a few sentances - the simple message is 2.4.X kernels will manage memmory resources on a low-memmory system better than a 2.0.X/2.2.X and the increase in kernel size is well worth it. Aside from performance issues the memomory management of the 2.4.X kernels also exhibit better security characteristics than early kernels.