2026-09-01 - stack read-ahead guard; obSCEne runs its whole suite under orbistoun
Next obSCEne wall (D444 oracle): 136-kernel/handoff
faulted at image+0x4502ff reading
0x600000801000. The handoff probe reads 20 words through
rdi (the argument block orbistoun builds at the stack top);
the block sat flush against the reserved top with only 0x80
bytes of headroom, so a 20-word (0xA0) read ran one word
past the mapped stack. Fixed with a readable guard page above
the initial stack pointer (GuestStack
READAHEAD_GUARD), mirroring the unmapped overflow guard
below - a modest over-read of the handoff now lands on mapped zeroes, as
on hardware. Unit test reads both ends of the guard back as zero.
Result: obSCEne now runs all 555 checks under orbistoun and
reaches OBS|end (was crashing ~check 40). Tally
520/10/7/18 vs console 241/79/192/43. The ten
failures are enumerated in D445 as the next targets:
virtual-query-stack/-text (virtual_query blind
to image/stack - one fix, two checks), 900-surface/control
(resolver reports a non-existent symbol present),
110-modules (module-list gap),
135-sysctl/osrelease and
137-kernelcall/system-version (refused; measured on
hardware, answerable from the report). mem tests pass (23), clippy
clean, fmt-clean.