Open Source Summit + ELC Europe 2019

Normally, memory allocations in kernel are done using kmalloc(), vmalloc() or their hybrid kvmalloc() functions. They all are baked by the page allocator and alloc_page() functions family. But during system boot when the most of the platform initialization code is run neither of them is available because the allocators are not yet set up.

For early memory management Linux has an allocator called `memblock` that provides the earliest abstraction of the physical memory and can be used nearly from the very beginning of the kernel execution.

This talk will cover the memblock APIs, the expectations and requirements for the architecture specific parts of the memory management setup and will wrap up with description of page allocator initialization.

HyperBus is a high performance 8-bit Double Data Rate bus used to connect SoCs with high performance flash devices. HyperFlash is a NOR based, Common Flash Interface (CFI) compliant HyperBus memory device. It’s increasingly replacing Parallel flashes due to faster boot time and reduced pin count. In this presentation, Vignesh will talk about the HyperBus framework he introduced recently in the Linux kernel under Memory Technology Devices (MTD) subsystem and how it supports HyperBus memory devices like HyperFlash.
The presentation introduces HyperBus protocol, CFI specification, HyperFlash and their kernel framework. It also provides an overview of how to write a new HyperBus Memory Controller Driver.
Finally the presentation talks about improvements made to CFI framework to reuse existing code in supporting HyperFlash and the challenges and problems that still need to be addressed.