[Skiboot] [RFC PATCH v7 00/12] Ultravisor support in skiboot

[Oliver O'Halloran]

On Tue, Jul 7, 2020 at 2:15 AM Ryan Grimm <grimm at linux.ibm.com> wrote:
>
> Oliver,
>
> These are the latest and greatest patches which include feedback from Alexy.
> They've gotten a decent amount of testing on Witherspoon and Swift.
>
> Since the last version, the ucall calling convention has been fixed up.  While
> I was at it I made sure the functions in ultravisor.h have nice comments.
> Also, region used by Mambo and BML which contains the ultravisor firmware is
> reserved properly.
>
> I want to write a unit test for local_alloc and local_free, but, I spent a lot
> of time and got stuck because I don't know what I'm doing.  I'll start a thread
> about designing this test.

You could probably re-use or modify some of the existing allocator
tests. What were you thinking of doing?

> My last TODO was look into probing the partition to avoid passing
> UV_LOAD_MAX_SIZE, but, I don't it's worth pulling in that probing code into
> hdata/spira.c, don't you think it's OK to be lazy and use a sane define?  Or
> would we want a ffs_part_get_sz in libffs.h that takes a name and returns a
> size?

Is there any reason we need to create the UV node in the HDAT parser?
We could defer creating it until after we've initialized flash. It
doesn't appear to be looking at anything from HDAT, just for the
presence of a secure-memory node.