Kido phase-2 proof — QEMU runs, 2026-10-18 Toolchain: aarch64-linux-gnu-gcc 13.3.0, qemu-system-aarch64. Harness: QEMU virt / cortex-a57 / 128 MiB, raw binary via -kernel (x0 = DTB), `timeout 8`, `-d int -D int.log`. Entry objects first in the link line (session-11 lesson); readelf-verified _start == _image_start == load address on both images. Pristine scaffold sources; only comment edits (proof-status headers). ---------------------------------------------------------------- RUN 1 — ACCIDENTAL relocated run (delta = +0x80000) ---------------------------------------------------------------- Harness error (owned): the linker-script copy's sed replaced RAM_BASE = 0x40000000 with itself — the canonical script was misread as already-0x40080000. Image linked at 0x40000000, QEMU -kernel loaded it at 0x40080000. The planned "control" was accidentally the relocated experiment. Evidence below is real and is labeled accidental. Console: SILENT (no output). kido_halt() spins without printing; the tripwire refused before kido_boot()'s breath print. The signal is breath-vs-silence against Run 2. -d int log: 0 bytes (zero exceptions). Exit: 124 (timeout in the designed halt spin). Register trace (-d cpu), first _start block (PC=0x40080024): X09 = 0x400836c0 runtime(probe) — PC-relative adrp/add, actual X10 = 0x400036c0 linked(probe) — literal-pool absolute X20 = 0x00080000 delta = +0x80000, the TRUE relocation X19 = 0x44000000 DTB pointer stashed PSTATE: EL1h (already at EL1; went straight to el1_entry) Branch trace: _start → el1_entry → BSS loop → stash → kido_entry_c_main → kido_reloc_verify → detect → crosscheck (agreement branch TAKEN: both instruments measured +0x80000) → delta policy (delta != 0 → KIDO_ERR_EARLY_ENV) → kido_halt spin. Proves: the .S measured the true delta in registers before SP existed; SP = _stack_top + delta kept the machine alive (no exception storm); the C instrument independently agreed; the tripwire refused LOUD by design. Session-12's declared exposure (relocated image dying before the tripwire) closed empirically. Also observed: the BSS loop zeroed the LINKED region (x0=0x40004618) — the known non-PI exposure; execution refused, so harmless; phase 3's problem. ---------------------------------------------------------------- RUN 2 — DELIBERATE delta-0 control (linked at load, 0x40080000) ---------------------------------------------------------------- Harness linker copy: single-line diff vs canonical, RAM_BASE = 0x40000000 → RAM_BASE = 0x40080000 (diff-verified) readelf: _start == _image_start == 0x40080000. Console: [Kido] outer kernel, first adaptive breath [Kido] arch: arm64 | handing off to kido_main... [Kido] kido_main: scheduler init [Kido] kido_main: 3 tasks created [Kido] kido_main: entering scheduler (never returns) ABCABCABCABC -d int log: 0 bytes (zero exceptions). Exit: 124 (timeout in the designed loud-halt spin on the emptied queue — expected). Proves: the CURRENT tree (phase-2 .S, reloc verify wiring, kido_image_actual_bounds, shifted allocator carve — all changed since session 11) boots clean on bare metal; the tripwire's delta-0 path (measure 0, crosscheck agree, proceed) works; the assembled context switch executes on allocator-backed stacks with the image bounds carved out (12-mark perfect interleave, zero exceptions = no stack/image overlap). ---------------------------------------------------------------- Regression gate (hosted, before QEMU — all real, gcc 13.3.0, -std=c11 -Wall -Wextra -Werror, zero warnings): reloc 21/21 | image 15/15 | alloc 31/31 | sched 75/75 | resonate 59/59 | interop 75/75 | mem ALL PASS | hal PASS | entry PASS | boot integration PASS (kido_boot() -> 0, AAAABBBBCCCC)