Conversation

Julian Andres Klode 🏳️‍🌈

I couldn't talk about this publicly before, but I reported a vulnerability in systemd-boot where it loads unsigned dtbs, if you sign systemd-boot yourself for internal uses, on arm64 or riscv64, you want to follow the advisory in

https://github.com/systemd/systemd/security/advisories/GHSA-6m6p-rjcq-334c

and apply the patches.

1
4
0
@juliank I'm not sure how to post on that issue, but I also don't know of any RISC-V systems that support this secure boot flavor (or really any proper secure boot, for that matter). It's not my area of expertise so I might be missing something, but IIUC there's still a bunch of work we'd need to make this all fit together.
1
0
0

@palmer Oh I think it is possible to build u-boot for various platforms with EBBR secure booting thingies, but I don't know.

But these are the only two cases where there could theoretically be problems.

1
0
0
@juliank Makes sense, there's at least arm64 hardware that has the necessary bits to be secure by the time it gets to u-boot.

I might be wrong here, but I don't know of any RISC-V hardware that can be configured to verify the first code that gets loaded from off chip. So as long as the thread model allows for messing around with a SPI flash (or wherever that code is loaded from), then we've got other problems that would require more of a HW-oriented fix.

Still a good fix and all, as we'll get there eventually and landing these backports in time can be a ton of work.
2
0
0

@palmer secure boot threat model really only targets UEFI rootkits placed in your storage from inside your OS.

1
0
0

@juliank @palmer err... Well, I'd have said rootkits on a bootable device, placed by any means. But notably "bootable device" here means not the firmware volume and not the EFI variable store.

1
0
0

@vathpela yeah it's hard to describe, but a local attacker usually has better means of attack, install keylogger hardware or shit inside the device.

@palmer

0
0
0

@palmer @juliank I listed riscv64 together with arm64 because I know it's possible to at least build sd-boot for it (we do it in Debian), so there's affected software out there. Whether it can actually be used in practice in the real world is anyone's guess, and that's part of the reason it's low severity

0
0
0