Jessica Clarke

https://github.com/CTSRD-CHERI/cheribsd/pull/1517 probably fixes whatever this "new problem that [you] don't really understand" was

> * Don't ask for a hostname This seems like it's asking for trouble... > * Shorten the pure-capability kernel name so it fits in the dialog Not a problem...

Oh, hm, presumably that means the bug still stands then :(

The final commit is going to need a lot of work. That breaks freebsd32 (maybe we don't care) but, more importantly, isn't deriving capabilities anywhere and just casting things around...

The IOMMU patch really needs to separate out the CHERI changes from the functional upstreamable IOMMU changes (that should be upstreamed), right now it's hard to review as a result...

You still have fixup commits for earlier commits in this PR

There is the question of whether libproc seeing the PC as the architectural PC or the address is the right choice. In either case, why are we even querying the...

I don't like this, and view it as a feature not a bug that we ditched all the bogus debugging crud from jdw57_memcpy.c (both the abort and the warning message)....

But regardless, this belongs in a common C file or header that jdw57_memcpy can share with it, as well as to keep bcopy.c as untainted as possible.

Also, this won't help you if/when we add hardware support for unaligned capability memcpy, since those _must_ silently strip tags and so you won't hit the memcpy case.