Intel Sends In More Habana Labs Gaudi2 AI Accelerator Code For Linux 6.1

Last cycle with Linux 6.0 initial Gaudi2 support was merged to the "habanalabs" AI driver that originally was focused on the Gaudi and Goya accelerators. With Linux 6.1 more of that Gaudi2 support has been prepped.
For that second-generation training and inference accelerator there are many driver additions including a user-space API for reading device attestation data, the uAPI to control the running mode of the engine cores, improved logging in case of hardware error events, improved firmware interactions, and other hardware enablement work.
At the same time as announcing Gaudi2, Intel also announced the Habana Labs Greco accelerator but so far that work hasn't seen any direct mainline activity. This pull request outlines the char/misc changes in full for Linux 6.1. There are also many IIO driver updates, FPGA subsystem updates, MHI changes, Coresight driver updates, and other random changes throughout.
Add A Comment