2026-08-31 - obSCEne's hardware run absorbed: five vaddrs confirmed, and a payload retry
obSCEne's 139-exports ran as a payload on the console
and confirmed seven of eight candidate export vaddrs by behaviour
(D410). Promoted the five newly-confirmed ones - getuid
0x630, geteuid 0x650, getgid
0x870, getppid 0x7d0, sceKernelGetProcessTime
0x16160 - from candidate to confirmed in
data/libkernel-vaddrs.txt;
libkernel_provenance now reports them Confirmed.
sceKernelGetTscFrequency 0x1cf30 was refuted but stays a
candidate: the offset is a real export-table placement, and what failed
was the probe's no-arg-getter assumption, not the address. Removed a
duplicate getpid 0x5b0 confirmed line.
Retried payload support against prosperous/klog.elf,
which exercised the whole path end to end. Plain run enters
with argc in rdi (elfldr's "call address 1") and dies at
0x1. Under --profile prospero-cex-12.40
(firmware skeleton maps at 0xf000000000, base handed at
0xf040000000) plus
ORBISTOUN_ENTRY_ARGUMENT=handoff, the payload reads
payload_args[0]=getpid, computes its base, and issues seven
syscalls - getpid (20) and vendor_system_version (649) through the
getpid+10 gadget - before hitting a new wall: an illegal instruction at
image+0x2708, a computed jump from a syscall return. The
fault moved from 0x1 to image+0x2708. As
expected, the provenance edits are metadata and did not change how far
the payload got; this was a no-regression confirmation, and orbistoun
correctly declined to record it as title progress because it ran under
the handoff diagnostic.
Two build notes for the next session. The workspace does not compile
on Linux/WSL right now: orbistoun-mem calls
rustix::param::page_size() but the workspace rustix
dependency enables only mm and thread, not
param (Cargo.toml is mid-edit, AM, so this was
left for whoever is editing deps). The rustix::param call
is under #[cfg(unix)], so the native Windows build is
unaffected - that is where this run was built and executed. Second,
--profile prints "ORBISTOUN_MACHINE_PROFILE is not a
diagnostic this build understands - ignored" yet the firmware still
maps, so the profile takes effect through the service layer while the
env passthrough warns; worth reconciling but not blocking.