OSDev.org

Please excuse my ignorance.
Suppose you run an i386 in protected mode with a virtual memory manager kernel such as Linux.
Suppose further that the unlikely condition that the ram is massively fragmented to the point of having 4K block used, 4K block free, 4K block used...etc
If a programs allocates 8K of memory and writes to all of it, how does the CPU know to skip over the used block? I did not think the CPU had a concept of nonlinear memory and I did not know there was a mechanism to intercept memory access other then page faults. How are virtual memory managers implemented?

smeezekitty wrote:I did not think the CPU had a concept of nonlinear memory

Yes, the CPU designer knows that, and therefore paging.

When paging is enabled, the machine (mostly) operate on linear address (aka logical address space),
an MMU translate that linear address into physical address using page table

For simplicity, we define: So for example, It's legal to have MMU(0) = 0 and MMU(4096) = 8192,
In this case physical memory addressed by 0 to 8192 is non-continuous.

Ahhh... I see a forgot about the Memory Management Unit
I cannot seem to find technical docs on the programming information and general specs.

Check Intel manuals volume 3 (System Programming). They have full description on how address is translated in different modes.