There will be a simulator later this year. I expect CheriBSD on Morello will released at the same time with a feature set compariable with MIPS and RISC-V.
The board seems much less interesting than this feature it delivers. "CHERI" sounds really cool. At first I thought it was a rehash of pointer auth/hw sanitizer ARM features that already exist. But no [1]:
> A Portable Architectural Protection Model ... We have experimented with adding the CHERI protection model to a number of ISAs, beginning with 64-bit MIPS, which was the baseline for our original research. More recently, we have developed an experimental version for 64-bit ARMv8-A (in collaboration with Arm), and also for 32/64-bit RISC-V.
Something that can work across architectures will see much faster adoption, I'm happy to see it.
We (University of Cambridge) have added it to MIPS and RISC-V. As the same team did both of these they feel similar. Arm is working closely with us to add it to their architecture, but it has a few differences, mostly to allow for experimentation to see what should be added to the future production architecture.
If I understand correctly, this is going to be the first custom silicon implementation of a CHERI-style capability architecture. So far the CHERI-MIPS and CHERI-RISC-V implementations have been FPGA-only.
That is correct. We have FPGAs, qemu, and formal models (that can boot an OS) of CHERI-RISC-V and CHERI-MIPS. Most software development uses qemu as it is faster than the FPGAs at around 100MHz.
Having something in the GHz range will give us a chance to try things we are currently struggling with, e.g. fuzzing is painfully slow.
8 comments
[ 2.9 ms ] story [ 28.8 ms ] threadprototype board to be available from late 2021
I didn't realize the wait would be quite that long.
> A Portable Architectural Protection Model ... We have experimented with adding the CHERI protection model to a number of ISAs, beginning with 64-bit MIPS, which was the baseline for our original research. More recently, we have developed an experimental version for 64-bit ARMv8-A (in collaboration with Arm), and also for 32/64-bit RISC-V.
Something that can work across architectures will see much faster adoption, I'm happy to see it.
[1] https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-941.pdf
Having something in the GHz range will give us a chance to try things we are currently struggling with, e.g. fuzzing is painfully slow.