Abstract :
In operating system design, efficiency often comes down to a simple philosophy:
Never do today what you can put off until tomorrow. In this talk, I will share my experience that I’ve gained by tinkering with the Linux Kernel and implementing things such as - Demand Paging and Lazy Loading within Aurora, a custom kernel I am building from scratch for XenevaOS.
I will walk through how Aurora manages large executables by mapping them incrementally into virtual memory, using processor traps as the trigger to allocate physical frames only when accessed. A key focus will be the architectural challenge of tracking page metadata. I’ll discuss why I moved beyond simple virtual page marking to implement a dedicated physical page database - drawing direct inspiration from how the Linux kernel utilizes struct page for its 4KiB frames.
By comparing my implementation of Copy-On-Write (COW) and file-backed page tracking in Aurora with established Linux paradigms, I aim to provide a "ground-up" perspective on memory management. This session is for anyone interested in how theoretical OS concepts translate into functional code and the practical trade-offs involved in building a memory subsystem from the floor up.
--
You received this message because you are subscribed to the Google Groups "Kernel Meetup Bangalore" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kernel-meetup-ban...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/kernel-meetup-bangalore/bce60684-d623-49b2-bc91-56284d11f8ban%40googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/kernel-meetup-bangalore/D4694209-FFA7-4C1D-BEDC-7FBB45B97FDC%40gmail.com.