On Mon, 2019-05-20 at 05:32 +0000, Wu, Hao A wrote:
> My thought on this is the CSM components in the framework packages are
> stable for a period of time. So my guess is that the issue is not directly
> related to these components.

Right. I went back to a SeaBIOS and OVMF from the time I last had this
working, and then I tripped over lots of toolchain issues. I ended up
having to git-bisect for various different issues at once.

ISTR I eventually got to the point where I could build the "last known
good" versions with the minimal set of fixes... and still they didn't
actually work. I need to revisit that.

> Also, if it turns out that there are some missing components left
> uncopied, we are able to get them back (from the repo history) and put
> them into OvmfPkg then. I can help on that.
> 
> Does this sound good to you?

That would be great. Thanks. I don't think there's anything really
*missing*. We do get into SeaBIOS as a CSM but SeaBIOS itself then
crashes somehow, in a hard-to-debug way.

In fact I did all this before we had proper SMM support in OVMF and
SeaBIOS, and I should probably revisit it completely. Originally I took
the simple approach where SeaBIOS takes over the hardware completely,
and returning from CSM to UEFI on a boot failure was not really going
to work. But if I move to what I understand is the "normal" CSM model
of invoking UEFI services through SMM instead of taking full control,
things might be a little saner.