On Mon, Jun 03, 2024 at 04:31:45PM +0200, Paolo Bonzini wrote: > On Mon, Jun 3, 2024 at 4:28 PM Michael Roth <michael.roth@xxxxxxx> wrote: > > So for now maybe we should plan to drop it from qemu-coco-queue and > > focus on the stateless builds for the initial code merge. > > Yes, I included it in qemu-coco-queue to ensure that other things > didn't break split firmware (or they were properly identified), but > basically everything else in qemu-coco-queue is ready for merge. > > Please double check "i386/sev: Allow measured direct kernel boot on > SNP" as well, as I did some reorganization of the code into a new > class method for sev-guest and sev-snp-guest objects. The patch changes look sensible to me, and I re-tested the following permutations of split/unsplit OVMF with/without kernel hashing and everything looks good (this is with PATCH 29/31 reverted): |split |split |unsplit |unsplit | |hashing=on|hashing=off|hashing=on|hashing=off| |----------|-----------|----------|-----------| svm | n/a | PASS | n/a | PASS | sev | n/a | PASS | PASS | PASS | sev-es | n/a | PASS | PASS | PASS | snp | n/a | n/a | PASS | PASS | (split + hashing=on is not possible because hashing requires AmdSevX64 OVMF package which is built unsplit-only by design. split tests done using OvmfPkgX64) -Mike > > Paolo >