Hi All, Please see Joakim's notes from his recent OSFC attendance below. Thanks for attending and sharing the notes Joakim. :) Best, Don
---------- Forwarded message --------- From: Joakim Bech via Board board@lists.trustedfirmware.org Date: Mon, 3 Oct 2022 at 01:20 Subject: [Board] OSFC 2022 To: board@lists.trustedfirmware.org, Okash Khawaja okash@google.com
Hi,
From Linaro, Ilias and I were attending the Open Source Firmware Conference (https://www.osfc.io) in Gothenburg in September. As you remember TrustedFirmware decided to sponsor the event again, so I was actually attending on behalf of TrustedFirmware. I think it was a great event, one of the best I've been at I think.
A common theme seemed to be "less is more", i.e., my impression after listening to the talks and after having discussions with people, it feels like people believe that various projects have had a bit too much feature creep (BMC's and EDK2 was brought up a couple of times as an example). Another issue is the slow response time on getting things fixed in BMCs, Management Engines etc. On average it took 18 months to get reported (security) issues fixed. Related was the complexity of having a lot of other code running outside the main OS (again ME's, BMC's, dedicated security blocks etc). General impression was that people would like to get back into more controlled environments.
DICE [1] (RIoT) from TCG/Microsoft seems to be getting more attention and it's starting to find its way into more devices. Recently we've heard this being mentioned by a few independent companies as a possible and simple lightweight solution to devices in need for some device identity and to be able to do some measure boot without having to rely on a TPM device. We (Ilias) presented DICE to the Linaro LEDGE group half a year ago as a potential area of interest. We'll bring this up again to a greater audience at Linaro and eventually we'll propose something that will affect TF-A . The DICE engine could run in BL1 and the DICE core could live in BL2. If that discussion matures, we'll have to bring it up to the TF TSC as well.
[1] https://www.microsoft.com/en-us/research/project/dice-device-identifier-comp...
There still seems to be a misconception about UEFI, that UEFI == EKD2. To some extent I believe that we were able to communicate that U-Boot contains tiny subset of UEFI, making it possible to boot EFI and that our end goal with ongoing Linaro work is to make it possible to boot any Linux distro (and possible also Windows) without having to make devices/platform specific changes to the OS side. On this matter, we've also synced up with Simon Glass at Google. As you know Simon is proposing an alternative implementation called VBE, which has a different approach. In some sense it's a cleaner and more simple solution, but we believe it will be hard to reach the goal of running any distro without relying on device specific customizations when using VBE.
Google (Thordur Bjornsson), mentioned challenges with attesting hardware on the upcoming v9 (CCA). He claimed that Intel did that part right, although the security solution around it later on was broken. I think we should introduce him to Charles Garcia Tobin.
I briefly had a chat with Christian Walter (9element) who is one of the OSFC organizers. He was grateful that TrustedFirmware sponsored the event again and that we seemed to like their event.
Mullvad (Swedish VPN provider and also sponsor) released a new USB key called tilitiskey [2]. They gave a demo where they authenticated a user for a SSH session. Their solution is kind of built using DICE as well (they mix in additional user provided data as well into the hash). We all got engineering samples, it should be fun to see how that project turns out.
@Okash, perhaps you have something to share as well?