QEMU GIC SGI proof, 2026-10-25 (scratch harness, /tmp only — not a module) Command: qemu-system-aarch64 -machine virt,gic-version=2 -cpu cortex-a57 -m 128 -nographic -kernel proof.bin -d int (proof.bin: scratch vec.S + main.c + committed modules/kido-gic.c, linked at 0x40080000, entry _start == load address) Console output: SKIG S = scratch_main start K = kido_gic_init OK (dist 0x08000000, cpu 0x08010000) I = IRQ unmasked (DAIF), about to send SGI G = SGI 0 handler ran (via kido_gic_dispatch from the IRQ stub) Exit 124 = timeout in the designed post-proof spin loop (expected). -d int log (202 bytes, full contents): Taking exception 5 [IRQ] on CPU 0 ...from EL1 to EL1 ...with ESR 0x0/0x3800000 ...with ELR 0x40081138 ...to EL1 PC 0x40080a80 PSTATE 0x3c5 Exception return from AArch64 EL1 to AArch64 EL1 PC 0x40081138 Reading: exactly one IRQ taken, one clean exception return. No fault-class entries (no sync/data-abort/undefined-instruction). The single G (not a storm) proves EOIR deactivated the SGI.