I was under the impression that a 32bit machine has only a 32bit addressrange of virtual memory which ALL processes share - but then via one memory address translation table per process, the process' "offset" in that vmem addressspace is stored, so that each process can act *as if* it has its own vmem?
I mean - how else can this work? If each process actually had its own 32bit vmem addresspace, then how would the kernel avoid collisions when assigning memory to processes? I mean, it someway needs to know which proc has which memory mapped, and surely it doesn't want to parse all processes address-translation tables? I dont understand anything anymore - please enlighten me
P.S.: Okay, after clearing my confused mind a bit, i think i can express my question more cleanly. Assuming that each proc has its own vmem, its address translation table would map ressources - also RAM - in that table. When applications request memory, they need to ask the kernel for it. However, regarding RAM-backed vmem pages, the kernel may not give two processes the same physical ram. But when the kernel is asked for mem, how does he know which ram-pages are used by which process? It cannot parse the vmem address translation tables of every process for every ram allocation - that would be too slow. So how does the kernel know? Does it keep its own central list of which procs use which ram pages?
I got some other questions, but the above one is the major missing puzzle piece to me.
