b38: S21 table-walker router

docs/sbl-bringup.md

@@ -41,3 +41,168 @@
 1. Добить баннер (см. свежий in_asm-хвост).
 2. Заменить стаблеты настоящим PBL (нужны дампы с живого тела!).
 3. UEFI (img3, EDK2) — отдельный фронт после SBL1.
+
+## b18: S7 и цикл delay-helper (2026-09-12)
+Механизм зацикливания (b17, доказан Unicorn-harness `tools/sbl_uni.py`):
+- sbl_main → delay-функция `0x8013B78` (256-итераций table-walk, `bl 0x8017AF4`
+  каждую итерацию) → b14 (`0x8017AFC: b.w 0x8005B9E`) → helper `0x8005BB0`
+  → выход `ldr pc,[sp],#0x14` (0x8005BCA) → S3-слот (0x8013BA5, рестарт).
+- Эпилог delay (`0x8013BAC: pop {r3-r7,pc}`) снимает 0x18 без push; каждый
+  S3-рестарт съедает 0x18 живого стека. Маршрут: слоты → fill (STUBV →
+  pop pc = maze) → pristine IMEM (нули → null-jump → hook_null → maze →
+  stale lr) → переполнение IMEM (0x8610008).
+- Heap SBL (stm в 0x8006684) затирает maze/S6/banner (0, 0x7a, STUBV).
+  QEMU: dstub-overlay `[0x08006B80,0x08006C50)` (read = снапшот ELF+стабы,
+  write = игнор). Uni: deferred-restore guard (MEM_WRITE-хук срабатывает
+  ДО записи, restore надо делать в hook_code!).
+- Maze-blob: `movs r0,#0; lsr ip,lr,#24; cmp #8; bne hang; bx lr`
+  (возврат 0 только для lr из DDRLOW 0x08..., иначе hang).
+Фикс b18: патч BAC (`f8bd0a4a` → `f3f730b8`, `b.w S7`) + S7-стаблет
+(52B @ 0x08006C10): `cmp r4,#0xff` → iterate (`0x8013BA5`) или done
+(restore r4/r5/sl/fp, возврат в sbl_main `0x802f681`). BAE-путь
+(`[r7]==2`) принесён в жертву (r7=0, недостижим).
+Фронтир b18: sbl_main возвращается, новый код `0x801Bxxx`
+(`strb [r7,#2]`-паттерн, b17-NOP на 0x801B71E), падеж pop'ом пустого
+стека у верха IMEM — следующий фронт.
+
+## b18 parity (QEMU == uni, 2026-09-12)
+- QEMU поправлен до паритета: dstub-overlay + vec-maze для Thumb
+  (null-fetch исполняет maze и возвращается) + убран висячий префикс
+  bl (bogus odd store -> alignment fault, DFAR=0x8006B8D).
+- QEMU b18 доходит до того же фронтира, что uni b18 (38961 инстр.):
+  `0x801Bxxx`, r1=0x80522A4, r2/r3=STUBV, lr=0x801B73F, sp марширует
+  к верху IMEM. Смерти чуть различаются по таймингу (uni: read
+  0x8610000; QEMU: null-jump → maze-hang c lr-top 0x80), регион тот же.
+- Следующий фронт (оба): лабиринт `0x801B6xx` (логгер, r7=0/r4=0,
+  pop пустого стека). Кандидаты: S8-стаблет, таблица логгера.
+
+## b21-b23: логгер, S9, S10, rollback (2026-09-12)
+- Логгер `0x8034810` = голый `pop.w 0x28` без push (жрал стек) → S8:
+  вход → `bx lr` (b21). Возврат игнорируется (callers ставят r0=0).
+- Лабиринт `0x801B6A0`: счётчик `[r4]++` + `pop.w` на пустых фреймах
+  (r4/r7=STUBV из fill) → S9 (b22): `[sp+0x14]` ненулевой → исходный
+  pop; ноль → restore r0/r4/r5/sl/fp + **reset sp=0x085FFF00** +
+  прыжок в sbl_main `0x802f69d` через ip (без литералов).
+- Rollback `boot_rollback_version.c:580` (query всегда `0x302E`) → NOP NOP
+  (b22, возврат игнорируется). Query зовёт `*table[0x080528D0]` (нули) →
+  фабрикация: poke STUBV (maze вернёт 0 = версии OK) (b23).
+- Rollback выходил боком в helper-mid (`pop{r4,pc}` без push) → S10 (b23):
+  `movw/movt ip + bx ip` в sbl_main `0x802f6a3`.
+- b23: 2M инстр. без падений (sp здоров), sbl_main идёт дальше
+  (`0x80164C4`-subtree → прыжок в OCIMEM-нули `0x222F84` — фронт PBL-данных).
+  QEMU-паритет: R15=0x223000, sp=0x85FFF00, lr=0x80164CB — там же.
+
+## b24: world-switch stub, clock S11 (2026-09-12)
+- Диспетчер (`0x80164C4`-subtree) зовёт ARM jump-table `0x8007158`
+  (`ldr pc`), та — в OCIMEM-трамплин world-switch `0x221EF8`
+  (ARM: save regs, `cps MON`, `mrc`, `ldr sp/pc` из PBL-таблиц
+  `[0x221F6C]=0x223620`/`[0x221F70]=0x223930` → вызов TZ; TZ не загружен).
+  Вход получается в Thumb → b24-стаб (4B `movs r0,#0; bx lr`):
+  возврат 0 в диспетчер (TZ deferred). Перезаписывает голову ARM'а
+  (восстановимо из ELF).
+- Дальше: clock-fn `0x800E960` (udiv-подсчёты) с голым выходом
+  `pop{r4,pc}` (без push, tail-chain через NOP-идиому) → S11 (b24):
+  вход → `bx lr` (lr = брошенный-но-валидный `0x80076CD`).
+- Poll `[0x73A100]` (lsls#29+bpl = проверка bit2!): SBL только читает →
+  фабрикация bit2=1 (b24 scaffold, обе стороны).
+- `blx #0x8006964` в файловые нули паддинга (PBL scratch) → S12v2 (b25
+  scaffold, ARM! т.к. blx-imm на чётный): preset ip=S16 + ACK bit1
+  `[0x73A100]` (side effect для poll#2) + `bx lr`. Литерал `[0x8006980]`.
+- Helper-pop `0x8007B66` читает scratch (счётчики 0-4 вместо pushed lr,
+  фреймы съедены null-циклами) → S16v2 (b25 smart): B66 (`pop` → `bx ip`,
+  2B) → S16 проверяет ip (odd + top 8 = legit pushed-lr → `bx ip`;
+  иначе default: restore r7, прыжок `0x8007779`). S12v3 грузит ip из
+  `[sp+4]` (pushed lr) + ACK bit1. B66-патч в ELF (b25), S12/S16 в scaffold.
+
+## b26: S17 safe-logger (2026-09-13)
+- S8 (`bx lr`) гасил весь init-лог (UART пуст by design) вслепую.
+  S17 (50B @0x08006CF0): печать r1-строки в UARTDM-TF, только если r1
+  в rodata `[0x804B800,0x8058000)` (wild-указатели скипаются), затем
+  pop-return (без leak'а, возврат игнорируется). Логгер `0x8034810`
+  (`pop.w` 4B) → `b.w S17`. Видимость прогресса вместо счёта инстр.
+
+## b27: poll re-init skip (2026-09-13)
+- Poll-fn `0x8007748`: `cbz r6` (r6=2 константа caller'а, никогда 0) →
+  вечный re-init (`movs r4,#0`, table loop) вместо poll#2-direct.
+  Патч 2B→2B в `b`. Биты 0,1,2 `[0x73A100]` pref-set (scaffold).
+- S16v3: `add sp,#8` перед `bx ip` (снимает helper-фрейм, пропущенный
+  B66 — un-leak стека вниз).
+- S16v4: отбраковка `ip==STUBV` (fill/maze — никогда legit return) →
+  default вместо возврата в maze (разрыв maze-B66 петли).
+- S12v4 (scaffold, без ELF): smart dispatcher в самом S12 (проверка
+  pushed-lr + ACK + возврат) — S16 больше не нужен в цепочке (остаётся
+  как fallback).
+- S12v5: +restore r7 (peripheral base, S16 dormant) прямо в S12.
+
+## b28: poll-epilogue march kill (2026-09-13)
+- Poll-fn входит mid-way (без prologue-push), выходит через
+  `add sp,#0x1C + pop.w 0x14` (+0x30/проход, пушей нет) → марш вверх
+  через fill в IMEM-нули → fault на верху. `add sp` → NOP (2B, b28).
+  Pop читает helper-фреймы (pushed-lr, валидные возвраты).
+
+## b29: S18 poll-epilogue router (2026-09-13)
+- b28-NOP убрал только +0x1C; `pop.w 0x14` без пушей всё равно марширует.
+  S18 (10B @0x08006D22): слепой прыжок в sbl_main `0x802F6A9`
+  (post-dispatcher). Эпилог `pop.w` (4B) → `b.w S18`.
+
+## b30: epilogue blx (2026-09-13)
+- sbl_main дошёл до `0x8016320` (logger-ish, грузит магию "SBL1"),
+  тот зовёт shared-эпилог `0x8007280` (Thumb!) через `blx`-imm
+  (чётный → ARM → fault на coprocessor-мусоре). Патч в глобальной
+  BLX2BL (`blx` → `bl`, как остальные). r7="SBL1", r0=SMEM — читает
+  настоящие таблицы!
+
+## b31: logger-epilogue NOP (2026-09-13)
+- `0x8016320` зовёт эпилог `0x8007280`, чей pop читает fill/STUBV →
+  maze → stale-lr в logger-mid → снова blx эпилог (голая взаимная
+  рекурсия). Вызов скипается (NOP NOP); sbl_main идёт дальше
+  `0x802F6B0` в свежий код.
+
+## b32: logger-tail b.w (2026-09-13)
+- 9-й вызывающий shared-эпилога (`0x8016342 blx`, пропущен статикой):
+  точный `b.w` (флип съезжает +2). r7="SBL1" магика из настоящих таблиц.
+
+## b33: partition wrap-check (2026-09-13)
+- Валидатор партиций (`adds; blo success`, line 0x1FD) ассёртит:
+  r0+r4 не враппаются (PBL-таблицы нулевые, r4 stale). `blo` → `b`
+  (2B). TODO(MainOS): настоящие PBL/SMEM partition tables.
+
+## b34: hang neutralize (2026-09-13)
+- Assert-hang `0x8018458` (`b .`) достигается напрямую через
+  maze-stale-lr (мимо патченного wrap-check). `b .` → `b success`
+  (2B, в success-путь).
+
+## b35: sbl_mc hangs (2026-09-13)
+- Мёртвая assert-fn `sbl_mc.c:320` (`0x802F9A0-b.`, `0x802F9B2-b.`)
+  достигается accident'ом (pop-garbage). Оба hang'а → S19
+  (рестарт post-logger init, bounded table work). 4B `b.w` каждый.
+- Урок: полный IMEM-poison ломает IMEM-стек (счётчики читают
+  0xFFFFFFFF вместо свежих нулей → underflow → wild-pop). Только
+  tripwire 4B на входе дрейфа `[0x8600000,0x8600004)`.
+
+## b36: assert-entry to S19 (2026-09-13)
+- sbl_mc assert-fn входит mid-way (`0x802F9A2`), минуя S19-hang'и.
+  Вход тоже в S19 (ест ldr + половину movw).
+
+## b37: S20 clock-dispatcher router (2026-09-13)
+- Голый clock-диспетчер (`0x800EB80`, без пушей) снимает `pop.w 0x30`
+  за проход → марш sp к верху IMEM. S20 (34B @0x08006D38):
+  `[sp+0x2C]` валиден → исходный pop; иначе r0=sl + re-run
+  (идемпотентная математика). Эпилог `pop.w` (4B) → `b.w S20`.
+
+## b38: S21 table-walker router (2026-09-13)
+- Голый table-walker (`0x8007916`, mid-входы без пушей) снимает
+  `pop.w 0x28`, читает saved-FP (`0x85FFEB0`) как pc → ARM fill-дрейф
+  (STM-clobbers!). S21 (32B @0x08006D5A): `[sp+0x24]` валиден →
+  исходный pop; иначе `b.w S19`. Эпилог `pop.w` (4B) → `b.w S21`.
+
+## b37: S20 clock-dispatcher router (2026-09-13)
+- Голый clock-диспетчер (`0x800EB80`, без пушей) снимает `pop.w 0x30`
+  за проход → марш sp к верху IMEM. S20 (34B @0x08006D38):
+  `[sp+0x2C]` валиден → исходный pop; иначе r0=sl + re-run
+  (идемпотентная математика). Эпилог `pop.w` (4B) → `b.w S20`.
+
+## maze v2 (2026-09-13)
+- Maze с чётным lr уходил в ARM-дрейф (STM-clobbers + IMEM fault).
+  Maze v2 (24B): возврат только при odd Thumb lr из DDR, иначе hang
+  громко. Залез на голову S6 (`[B9E,BA4)`, S6 dormant). Vec-maze тоже v2.

qemu/maze.S

@@ -1,20 +1,14 @@
     .syntax unified
     .cpu cortex-a7
     .thumb
-    @ maze entry logger: record lr/sp to mailbox 0x08006964, return 0.
-    @ Guard: if lr is outside known-good SBL/stub ranges, spin with
-    @ interrupts disabled instead of jumping into the void (post-mortem
-    @ via vec trap regs).
+    @ maze entry: return 0. Guard: bad lr (high byte != 0x08) spins
+    @ with IRQs off instead of jumping into the void.
     .text
     .global maze_stub
 maze_stub:
-    movw r12, #:lower16:0x08006964
-    movt r12, #:upper16:0x08006964
-    str lr, [r12]
-    str sp, [r12, #4]
     movs r0, #0
     lsr r12, lr, #24
-    cmp r12, #0x08
+    cmp r12, #8
     bne maze_hang
     bx lr
 maze_hang:

qemu/pbl.S

@@ -0,0 +1,22 @@
+    .syntax unified
+    .cpu cortex-a7
+    .thumb
+    @ Minimal PBL stub (position-independent): BLX to SBL1 entry (seg1),
+    @ then canonical regs + jump to sbl_main. Mirrors PBL call-then-jump.
+    .text
+    .global pbl_stub
+pbl_stub:
+    push {r4-r7, lr}
+    movw r12, #0x6b81
+    movt r12, #0x800
+    blx r12                   @ SBL1 entry (odd: stay Thumb)
+    movw r0, #0x0000
+    movt r0, #0x0022          @ r0 = pbl_shared 0x00220000
+    mov r5, r0
+    movw sl, #0x0000
+    movt sl, #0x0806          @ sl = table base 0x08060000
+    movw fp, #0x1000
+    movt fp, #0x0806          @ fp = 0x08061000
+    movw lr, #0xf63f
+    movt lr, #0x0802          @ lr = sbl_main odd (noreturn marker)
+    bx lr

qemu/run-sbl.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+# run-sbl.sh — запуск SBL1 на машине saimaa (пути постоянные, не /tmp).
+# Использование: ./qemu/run-sbl.sh [bXX] [extra-qemu-args...]
+#   bXX — вариант патча (по умолчанию b17)
+set -e
+cd "$(dirname "$0")/.."
+QEMU=~/qemu-src/qemu-8.2.2/build/qemu-system-arm
+B=${1:-b38}
+IMG=fw/img/emmc-real.img
+if [ ! -f "$IMG" ]; then
+  echo "no $IMG — сначала: python3 tools/ffu_extract.py fw/dl/RM1127_059X5R5_orig.ffu --emmc-user fw/img/user-real.bin"
+  echo "и: python3 tools/mkimage.py --user fw/img/user-real.bin $IMG"
+  exit 1
+fi
+shift 2>/dev/null || true
+# timeout для тестовых прогонов (0 = без ограничения)
+T=${T:-15}
+if [ "$T" != "0" ]; then
+  exec timeout "$T" "$QEMU" -M saimaa,sbl1="fw/img/patch/img0_${B}.elf",sblmain=0x802f63f \
+    -m 512 -display none -monitor none -serial file:uart.log \
+    -drive file="$IMG",if=sd,format=raw "$@" 2>&1
+fi
+exec "$QEMU" -M saimaa,sbl1="fw/img/patch/img0_${B}.elf",sblmain=0x802f63f \
+  -m 512 -display none -monitor none -serial file:uart.log \
+  -drive file="$IMG",if=sd,format=raw "$@" 2>&1

qemu/saimaa.c

@@ -35,6 +35,10 @@
 /* High DDRLOW for our stubs (SBL heap reuses 0x0806xxxx) */
 #define SAIMAA_DDRHI_BASE    0x08610000
 #define SAIMAA_DDRHI_SIZE    0x00FF0000 /* ..0x09600000 */
+/* Stubs carve-out inside DDRHI (SBL heap overwrites plain RAM):
+ * maze+S6+banner+mailbox live here, set read-only after implant. */
+#define SAIMAA_STUBS_BASE    0x08800000
+#define SAIMAA_STUBS_SIZE    0x00000300
 #define SAIMAA_IMEM_BASE     0x08600000
 #define SAIMAA_IMEM_SIZE     0x00010000
 #define SAIMAA_DDRH_A_BASE   0x80000000
@@ -64,10 +68,13 @@ struct SaimaaTmrState {
 
 static uint64_t saimaa_tmr_read(void *opaque, hwaddr off, unsigned size)
 {
+    /* Deterministic tick (+1/read): advances delay loops but keeps runs
+     * reproducible (qemu_clock would diverge run to run). */
+    static uint64_t tick;
     (void)opaque;
     (void)off;
     (void)size;
-    return (uint64_t)(qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) >> 12);
+    return ++tick;
 }
 
 static void saimaa_tmr_write(void *opaque, hwaddr off, uint64_t val,
@@ -94,6 +101,57 @@ static void saimaa_tmr_init(Object *obj)
     sysbus_init_mmio(dev, &s->iomem);
 }
 
+/* Stubs region handle for set_readonly after scaffold implant. */
+static MemoryRegion *saimaa_stubs_mr;
+
+/* DDRLOW stub trap [0x08006B80,0x08006D80): serves the scaffolded bytes
+ * (S/maze/S6/banner/S7/S9/S10/S16/S17/S18/S20/S21 stubs) on read/fetch,
+ * ignores SBL heap-clobber writes. Uni parity: deferred-restore guard
+ * in sbl_uni.py (wider: [0x080068C0,0x08006D80), also covers PBL stub).
+ * Snapshot is taken AFTER ELF load + all stub writes (readback).
+ * PBL stub [0x080068C0,0x080068F4) and the ELF-text gap below stay plain
+ * RAM: PBL runs once (if at all), gap holds real hot SBL code. */
+#define SAIMAA_DSTUB_BASE 0x08006B80
+#define SAIMAA_DSTUB_SIZE 0x00000200
+static uint8_t saimaa_dstub_mem[SAIMAA_DSTUB_SIZE];
+
+static uint64_t saimaa_dstub_read(void *opaque, hwaddr off, unsigned size)
+{
+    uint64_t v = 0;
+    unsigned i;
+    (void)opaque;
+    if (off >= SAIMAA_DSTUB_SIZE) {
+        return 0;
+    }
+    if (off + size > SAIMAA_DSTUB_SIZE) {
+        size = SAIMAA_DSTUB_SIZE - off;
+    }
+    for (i = 0; i < size; i++) {
+        v |= (uint64_t)saimaa_dstub_mem[off + i] << (8 * i);
+    }
+    return v;
+}
+
+static void saimaa_dstub_write(void *opaque, hwaddr off, uint64_t val,
+                               unsigned size)
+{
+    static int n = 0;
+    (void)opaque;
+    (void)val;
+    (void)size;
+    if (n < 8) {
+        fprintf(stderr, "saimaa-dstub: ignored heap write @0x%x\n",
+                SAIMAA_DSTUB_BASE + (unsigned)off);
+        n++;
+    }
+}
+
+static const MemoryRegionOps saimaa_dstub_ops = {
+    .read = saimaa_dstub_read,
+    .write = saimaa_dstub_write,
+    .endianness = DEVICE_NATIVE_ENDIAN,
+};
+
 /* ---- minimal MSM UARTDM stub: TX prints, status always ready ---- */
 #define TYPE_SAIMAA_VEC "saimaa-vec"
 #define TYPE_SAIMAA_UART "saimaa-uart"
@@ -109,21 +167,46 @@ struct SaimaaUartState {
 static int saimaa_vec_n;
 static uint64_t saimaa_vec_read(void *opaque, hwaddr off, unsigned size)
 {
-    if (saimaa_vec_n < 400) {
+    /* Maze blob v2 mirrored at [0,24) for THUMB fetches: null-jump
+     * fetches (bx r0=0, pop 0, ldr pc,=0) execute the maze and return
+     * via lr — uni parity (hook_null -> STUB). ARM-mode fetches (real
+     * exception entry) keep BKPT vectors below. Data probes elsewhere
+     * read 0. */
+    static const uint8_t maze[24] = {
+        0x00, 0x20, 0x4f, 0xea, 0x1e, 0x6c, 0xbc, 0xf1,
+        0x08, 0x0f, 0x03, 0xd1, 0x1e, 0xf0, 0x01, 0x0f,
+        0x00, 0xd0, 0x70, 0x47, 0x72, 0xb6, 0xfd, 0xe7,
+    };
+    {
+        ARMCPU *cpu = ARM_CPU(current_cpu);
+        if (cpu && cpu->env.thumb && off < sizeof(maze)) {
+            uint64_t v = 0;
+            unsigned i, n = size;
+            if (off + n > sizeof(maze)) {
+                n = sizeof(maze) - off;
+            }
+            for (i = 0; i < n; i++) {
+                v |= (uint64_t)maze[off + i] << (8 * i);
+            }
+            return v;
+        }
+    }    if (saimaa_vec_n < 400) {
         ARMCPU *cpu = ARM_CPU(current_cpu);
         uint64_t pc = cpu ? cpu->env.regs[15] : 0;
         int th = cpu ? cpu->env.thumb : 0;
-        uint32_t mb0 = 0, mb1 = 0;
-        if (saimaa_vec_n < 3) {
-            cpu_physical_memory_read(0x08006964, &mb0, 4);
-            cpu_physical_memory_read(0x08006968, &mb1, 4);
+        uint32_t dfar = 0, ifsr = 0, dfsr = 0;
+        if (cpu) {
+            /* CP15 DFAR/IFSR/DFSR for abort diagnosis */
+            dfar = cpu->env.cp15.dfar_s;
+            ifsr = (uint32_t)cpu->env.cp15.ifsr_s;
+            dfsr = (uint32_t)cpu->env.cp15.dfsr_s;
+            fprintf(stderr, " SCTLR=0x%lx", (unsigned long)cpu->env.cp15.sctlr_s);
         }
         fprintf(stderr, "saimaa-vec: %s off=0x%x pc=0x%lx",
                 size == 2 ? "fetch?" : "read",
                 (unsigned)off, (unsigned long)pc - (th ? 4 : 8));
-        if (saimaa_vec_n < 3) {
-            fprintf(stderr, " maze-mailbox: lr=0x%x sp=0x%x", mb0, mb1);
-        }
+        fprintf(stderr, " DFAR=0x%x IFSR=0x%x DFSR=0x%x", dfar, ifsr,
+                dfsr);
         if (cpu) {
             int b, r;
             fprintf(stderr, " regs=");
@@ -248,19 +331,27 @@ static void saimaa_machine_init(MachineState *machine)
         { "saimaa.rpm", SAIMAA_RPM_BASE, SAIMAA_RPM_SIZE },
         { "saimaa.ocimem", SAIMAA_OCIMEM_BASE, SAIMAA_OCIMEM_SIZE },
         { "saimaa.ddrlow", SAIMAA_DDRLOW_BASE, SAIMAA_DDRLOW_SIZE },
-        { "saimaa.ddrhi", SAIMAA_DDRHI_BASE, SAIMAA_DDRHI_SIZE },
+        { "saimaa.ddrhi_lo", SAIMAA_DDRHI_BASE,
+          SAIMAA_STUBS_BASE - SAIMAA_DDRHI_BASE },
+        { "saimaa.stubs", SAIMAA_STUBS_BASE, SAIMAA_STUBS_SIZE },
+        { "saimaa.ddrhi_hi", SAIMAA_STUBS_BASE + SAIMAA_STUBS_SIZE,
+          SAIMAA_DDRHI_BASE + SAIMAA_DDRHI_SIZE -
+          (SAIMAA_STUBS_BASE + SAIMAA_STUBS_SIZE) },
         { "saimaa.imem", SAIMAA_IMEM_BASE, SAIMAA_IMEM_SIZE },
         { "saimaa.ddrhigh_a", SAIMAA_DDRH_A_BASE, SAIMAA_DDRH_A_SIZE },
         { "saimaa.ddrhigh_b", SAIMAA_DDRH_B_BASE, SAIMAA_DDRH_B_SIZE },
         { "saimaa.mystrom", SAIMAA_MYSTROM_BASE, SAIMAA_MYSTROM_SIZE },
         { "saimaa.smem", SAIMAA_SMEM_BASE, SAIMAA_SMEM_SIZE },
     };
-    for (i = 0; i < 10; i++) {
+    for (i = 0; i < ARRAY_SIZE(rams); i++) {
         MemoryRegion *ram = g_new(MemoryRegion, 1);
         memory_region_init_ram(ram, NULL, rams[i].n, rams[i].s,
                                &error_fatal);
-	memory_region_add_subregion_overlap(sysmem, rams[i].b, ram,
+ 	memory_region_add_subregion_overlap(sysmem, rams[i].b, ram,
                                               1);
+        if (!strcmp(rams[i].n, "saimaa.stubs")) {
+            saimaa_stubs_mr = ram;
+        }
         if (!strcmp(rams[i].n, "saimaa.mystrom")) {
             /* Mystery handler ROM: ARM mov r0,#0; bx lr repeating,
              * so unknown blx targets return 0 instead of faulting. */
@@ -387,6 +478,12 @@ static void saimaa_machine_init(MachineState *machine)
             ARM_CPU(first_cpu)->env.banked_r13[b] = 0x085FFF00;
         }
         ARM_CPU(first_cpu)->env.regs[13] = 0x085FFF00;
+        /* Stale lr (never bl'd on synthetic paths) returns into maze
+         * (returns 0 = SBL "empty" convention) instead of jumping to 0. */
+        ARM_CPU(first_cpu)->env.regs[14] = 0x08006B8D;
+        for (int b = 0; b < 8; b++) {
+            ARM_CPU(first_cpu)->env.banked_r14[b] = 0x08006B8D;
+        }
         ARM_CPU(first_cpu)->env.regs[0] = 0x00220000; /* fake pbl_shared */
         /* sbl_main expects r5 = mem-table pointer (PBL trampoline does
          * mov r5, r0 before jumping here); point it at pbl_shared too. */
@@ -432,7 +529,10 @@ static void saimaa_machine_init(MachineState *machine)
                    tgt & ~1u, (tgt & 1) ? " thumb" : " arm",
                    (unsigned long)entry);
         } else {
-            cpu_set_pc(first_cpu, entry);
+            /* Default: start at PBL stub (faithful call-then-jump flow) */
+            cpu_set_pc(first_cpu, 0x080068C1);
+            printf("saimaa: PBL-stub start -> entry 0x%lx then sbl_main\n",
+                   (unsigned long)entry);
         }
         /* Boot scaffolding for SBL1 BringUp (WIP stub-driven PBL):
          * static continuation slots for shared-epilogue pops.
@@ -440,30 +540,31 @@ static void saimaa_machine_init(MachineState *machine)
          * (sp values from cpu traces, b10 image) */
         {
             uint32_t magic = 0x00000000; /* IMEM poll spins WHILE equal */
-            uint32_t S = 0x08006900, STUB = 0x08006910;  /* in-code zeros, safe from heap */
-            uint32_t STUBV = 0x08006911; /* odd: blx stays Thumb */
+            uint32_t S = 0x08006B80, STUB = 0x08006B8C;  /* dead PBL-entry code */
+            uint32_t STUBV = 0x08006B8D; /* odd: blx stays Thumb */
             uint32_t w;
             uint16_t h0 = 0x2000, h1 = 0x4770; /* movs r0,#0; bx lr */
             uint32_t v;
             /* pointer maze in owned RAM for SBL1 struct chains.
-             * STUB = mailbox logger + return-0: records lr/sp at entry
-             * to 0x08006964 for post-mortem, then movs r0,#0; bx lr. */
+             * STUB = minimal logger + return-0 (movs r0,#0; bx lr);
+             * mailbox at 0x08006BD0 records last (lr,sp). */
             w = S + 8;
             cpu_physical_memory_write(S, &w, 4);
             /* [struct+4] is another fn ptr (e.g. 0x801b72e ldr r3,[r0,#4]
-             * with r0 = maze S via POOLS redirect); point at STUB too */
-            cpu_physical_memory_write(S + 4, &STUB, 4);
-            cpu_physical_memory_write(S + 8, &STUB, 4);
+             * with r0 = maze S via POOLS redirect); point at STUB too.
+             * NOTE: ODD (STUBV) — blx to even would switch to ARM! */
+            cpu_physical_memory_write(S + 4, &STUBV, 4);
+            cpu_physical_memory_write(S + 8, &STUBV, 4);
             {
-                /* from qemu/maze.S: mailbox logger + lr-range guard.
-                 * Bad lr (not 0x08xxxxxx) spins with IRQs off instead of
-                 * jumping into the void. 36 bytes. */
+                /* Maze v2 (24B @STUB): movs r0,#0; return 0 via lr ONLY
+                 * if lr is odd Thumb in DDR (top 8); even lr (stale
+                 * mid-fn raw addr!) would ARM-drift (STM-clobbers +
+                 * IMEM fault), so hang loud instead. Overlaps S6 head
+                 * ([B9E,BA4), S6 dormant, documented below). */
                 static const uint8_t mb[] = {
-                    0x46, 0xf6, 0x64, 0x1c, 0xc0, 0xf6, 0x00, 0x0c,
-                    0xcc, 0xf8, 0x00, 0xe0, 0xcc, 0xf8, 0x04, 0xd0,
                     0x00, 0x20, 0x4f, 0xea, 0x1e, 0x6c, 0xbc, 0xf1,
-                    0x08, 0x0f, 0x00, 0xd1, 0x70, 0x47, 0x72, 0xb6,
-                    0xfd, 0xe7, 0x00, 0x00,
+                    0x08, 0x0f, 0x03, 0xd1, 0x1e, 0xf0, 0x01, 0x0f,
+                    0x00, 0xd0, 0x70, 0x47, 0x72, 0xb6, 0xfd, 0xe7,
                 };
                 size_t k;
                 for (k = 0; k < sizeof(mb); k += 4) {
@@ -476,24 +577,12 @@ static void saimaa_machine_init(MachineState *machine)
             cpu_physical_memory_write(0x087c29d18, &S, 4);
             cpu_physical_memory_write(0x08050b28, &S, 4);
             cpu_physical_memory_write(0x08070000, &S, 4);
-            /* S6 stublet at 0x08006934: rebuild popped frame regs,
-             * return to 0x8013ba5 (b14 redirects 0x8017afc pop here) */
-            {
-                static const uint8_t s6[] = {
-                    0xbd, 0xe8, 0xfc, 0x47, /* ldmia.w sp!,{r2-r9,sl,lr} */
-                    0x43, 0xf6, 0xa5, 0x3e, /* movw lr,#0x3ba5 */
-                    0xc0, 0xf6, 0x01, 0x0e, /* movt lr,#0x801 */
-                    0x70, 0x47,             /* bx lr */
-                };
-                size_t k;
-                for (k = 0; k < sizeof(s6); k += 4) {
-                    uint32_t word = 0;
-                    size_t n = sizeof(s6) - k < 4 ? sizeof(s6) - k : 4;
-                    __builtin_memcpy(&word, &s6[k], n);
-                    cpu_physical_memory_write(0x08006934 + k, &word, 4);
-                }
-            }
-            /* Banner stub at 0x08006942: print r0 string via UARTDM
+            /* S6 stublet DISABLED (maze v2 overlaps its head [B9E,BA4);
+             * S6 dormant, never entered (b14 goes to 0x8005B9E).
+             * Bytes kept for reference (would live at 0x08006B9E):
+             * ldmia.w sp!,{r2-r9,sl,lr}; movw lr,#0x3ba5;
+             * movt lr,#0x801; bx lr. */
+            /* Banner stub at 0x08006BAC: print r0 string via UARTDM
              * (b16 redirects 0x802f690 blx here). Built from qemu/banner.S:
              * push frame, print loop, ldr pc,=0x802f694 (callee noreturn).
              * 28 bytes total. */
@@ -510,18 +599,263 @@ static void saimaa_machine_init(MachineState *machine)
                     uint32_t word = 0;
                     size_t n = sizeof(bn) - k < 4 ? sizeof(bn) - k : 4;
                     __builtin_memcpy(&word, &bn[k], n);
-                    cpu_physical_memory_write(0x08006942 + k, &word, 4);
+                    cpu_physical_memory_write(0x08006BAC + k, &word, 4);
                 }
             }
-            /* stack zone defaults to plain-stub addr: every register
-             * popped from untouched stack becomes callable (returns 0). */
+            /* PBL stub at 0x080068C0 (52B, from qemu/pbl.S): BLX to SBL1
+             * entry, then canonical regs + jump to sbl_main. Used when
+             * sblmain prop is unset (default): faithful PBL call flow. */
             {
-                uint32_t fill = 0x08006911;
+                static const uint8_t pbl[] = {
+                    0xf0, 0xb5, 0x46, 0xf6, 0x81, 0x3c, 0xc0, 0xf6,
+                    0x00, 0x0c, 0xe0, 0x47, 0x40, 0xf2, 0x00, 0x00,
+                    0xc0, 0xf2, 0x22, 0x00, 0x05, 0x46, 0x40, 0xf2,
+                    0x00, 0x0a, 0xc0, 0xf6, 0x06, 0x0a, 0x41, 0xf2,
+                    0x00, 0x0b, 0xc0, 0xf6, 0x06, 0x0b, 0x4f, 0xf2,
+                    0x3f, 0x6e, 0xc0, 0xf6, 0x02, 0x0e, 0x70, 0x47,
+                    0x00, 0x00, 0x00, 0x00,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(pbl); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(pbl) - k < 4 ? sizeof(pbl) - k : 4;
+                    __builtin_memcpy(&word, &pbl[k], n);
+                    cpu_physical_memory_write(0x080068C0 + k, &word, 4);
+                }
+            }
+            /* S7 stublet at 0x08006C10 (52B, b18): delay-epilogue exit
+             * router. cmp r4,#0xff: iterate (r4<=0xff -> 0x8013ba5) or
+             * done (restore r4/r5/sl/fp, return to sbl_main 0x802f681).
+             * Assembled with keystone (see docs/sbl-bringup.md). */
+            {
+                static const uint8_t s7[] = {
+                    0xff, 0x2c, 0x13, 0xd9, 0x40, 0xf2, 0x00, 0x04,
+                    0xc0, 0xf2, 0x22, 0x04, 0x40, 0xf2, 0x00, 0x05,
+                    0xc0, 0xf2, 0x22, 0x05, 0x40, 0xf2, 0x00, 0x0a,
+                    0xc0, 0xf6, 0x06, 0x0a, 0x41, 0xf2, 0x00, 0x0b,
+                    0xc0, 0xf6, 0x06, 0x0b, 0xdf, 0xf8, 0x00, 0xf0,
+                    0x81, 0xf6, 0x02, 0x08, 0xdf, 0xf8, 0x00, 0xf0,
+                    0xa5, 0x3b, 0x01, 0x08,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s7); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s7) - k < 4 ? sizeof(s7) - k : 4;
+                    __builtin_memcpy(&word, &s7[k], n);
+                    cpu_physical_memory_write(0x08006C10 + k, &word, 4);
+                }
+            }
+            /* S9 stublet at 0x08006C48 (62B, b22): labyrinth-epilogue
+             * exit router (0x801B6BE pop.w -> b.w S9). Empty frames ->
+             * restore r0/r4/r5/sl/fp, reset sp to 0x085FFF00, jump to
+             * sbl_main 0x802f69d via ip. Assembled with keystone. */
+            {
+                static const uint8_t s9[] = {
+                    0x05, 0x9b, 0x0b, 0xb1, 0xbd, 0xe8, 0xf0, 0x81,
+                    0x00, 0x20, 0x40, 0xf2, 0x00, 0x04, 0xc0, 0xf2,
+                    0x22, 0x04, 0x40, 0xf2, 0x00, 0x05, 0xc0, 0xf2,
+                    0x22, 0x05, 0x40, 0xf2, 0x00, 0x0a, 0xc0, 0xf6,
+                    0x06, 0x0a, 0x41, 0xf2, 0x00, 0x0b, 0xc0, 0xf6,
+                    0x06, 0x0b, 0x4f, 0xf6, 0x00, 0x73, 0xc0, 0xf6,
+                    0x5f, 0x03, 0x9d, 0x46, 0x4f, 0xf2, 0x9d, 0x6c,
+                    0xc0, 0xf6, 0x02, 0x0c, 0x60, 0x47,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s9); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s9) - k < 4 ? sizeof(s9) - k : 4;
+                    __builtin_memcpy(&word, &s9[k], n);
+                    cpu_physical_memory_write(0x08006C48 + k, &word, 4);
+                }
+            }
+            /* S10 stublet at 0x08006C88 (10B, b23): rollback-exit router.
+             * movw/movt ip + bx ip to sbl_main 0x802f6a3. */
+            {
+                static const uint8_t s10[] = {
+                    0x4f, 0xf2, 0xa3, 0x6c, 0xc0, 0xf6, 0x02, 0x0c,
+                    0x60, 0x47,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s10); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s10) - k < 4 ? sizeof(s10) - k : 4;
+                    __builtin_memcpy(&word, &s10[k], n);
+                    cpu_physical_memory_write(0x08006C88 + k, &word, 4);
+                }
+            }
+            /* S18 stublet at 0x08006D22 (10B, b29): poll-epilogue router.
+             * movw/movt ip + bx ip to sbl_main 0x802f6a9. */
+            {
+                static const uint8_t s18[] = {
+                    0x4f, 0xf2, 0xa9, 0x6c, 0xc0, 0xf6, 0x02, 0x0c,
+                    0x60, 0x47,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s18); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s18) - k < 4 ? sizeof(s18) - k : 4;
+                    __builtin_memcpy(&word, &s18[k], n);
+                    cpu_physical_memory_write(0x08006D22 + k, &word, 4);
+                }
+            }
+            /* S21 stublet at 0x08006D5A (32B, b38): table-walker router.
+             * [sp+0x24] valid-code? => original pop.w (legit return);
+             * else b.w S19 (sbl_main post-logger restart). */
+            {
+                static const uint8_t s21[] = {
+                    0x09, 0x9b, 0x13, 0xf0, 0x01, 0x0f, 0x09, 0xd0,
+                    0x1a, 0x0e, 0x08, 0x2a, 0x01, 0xd1, 0xbd, 0xe8,
+                    0xf0, 0x9f, 0x1a, 0x0f, 0x08, 0x2a, 0x01, 0xd1,
+                    0xbd, 0xe8, 0xf0, 0x9f, 0xff, 0xf7, 0xd9, 0xbf,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s21); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s21) - k < 4 ? sizeof(s21) - k : 4;
+                    __builtin_memcpy(&word, &s21[k], n);
+                    cpu_physical_memory_write(0x08006D5A + k, &word, 4);
+                }
+            }
+            /* S20 stublet at 0x08006D38 (34B, b37): clock-dispatcher
+             * router. [sp+0x2C] valid-code? => original pop.w (legit);
+             * else r0=sl + b.w dispatcher re-run (idempotent math). */
+            {
+                static const uint8_t s20[] = {
+                    0x0b, 0x9b, 0x13, 0xf0, 0x01, 0x0f, 0x09, 0xd0,
+                    0x1a, 0x0e, 0x08, 0x2a, 0x01, 0xd1, 0xbd, 0xe8,
+                    0xfe, 0x8f, 0x1a, 0x0f, 0x08, 0x2a, 0x01, 0xd1,
+                    0xbd, 0xe8, 0xfe, 0x8f, 0x50, 0x46, 0x07, 0xf0,
+                    0x13, 0xbf,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s20); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s20) - k < 4 ? sizeof(s20) - k : 4;
+                    __builtin_memcpy(&word, &s20[k], n);
+                    cpu_physical_memory_write(0x08006D38 + k, &word, 4);
+                }
+            }
+            /* S17 safe-logger (50B @0x08006CF0, b26): print r1-string
+             * via UARTDM-TF iff r1 in rodata [0x804B800,0x8058000);
+             * pop-return (no leak). Replaces S8 silence with logs. */
+            {
+                static const uint8_t s17[] = {
+                    0x4b, 0xf6, 0x00, 0x03, 0xc0, 0xf6, 0x04, 0x03,
+                    0x99, 0x42, 0x11, 0xd3, 0x48, 0xf2, 0x00, 0x03,
+                    0xc0, 0xf6, 0x05, 0x03, 0x99, 0x42, 0x0b, 0xd2,
+                    0x10, 0xb5, 0x4f, 0xf2, 0x00, 0x03, 0xc0, 0xf2,
+                    0x8a, 0x73, 0x0c, 0x46, 0x14, 0xf8, 0x01, 0x0b,
+                    0x08, 0xb1, 0x18, 0x67, 0xfa, 0xe7, 0x10, 0xbd,
+                    0x70, 0x47,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s17); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s17) - k < 4 ? sizeof(s17) - k : 4;
+                    __builtin_memcpy(&word, &s17[k], n);
+                    cpu_physical_memory_write(0x08006CF0 + k, &word, 4);
+                }
+            }
+            /* PBL version-table fabrication (b23): [0x080528D0] = STUBV
+             * so the rollback query's blx calls maze (0 = versions OK). */
+            {
+                uint32_t v = 0x08006B8D;
+                cpu_physical_memory_write(0x080528D0, &v, 4);
+            }
+            /* PBL dispatcher-table fabrication (b33, moved b37): retarget
+             * literal [0x80164F0] (image: hash blob!) to scratch struct
+             * at 0x08024000 (OCIMEM, out of stack-march paths!) with
+             * [0x18]=2 (ready -> dispatch). Was IMEM 0x08600200
+             * (self-executed when sp marched there!). */
+            {
+                uint32_t v = 0x08024000;
+                cpu_physical_memory_write(0x08024000, &v, 4);
+                v = 0x00000002;
+                cpu_physical_memory_write(0x08024014, &v, 4);
+                v = 0x08024000;
+                cpu_physical_memory_write(0x080164F0, &v, 4);
+            }
+            /* HW-ready bits (b25/b26): [0x073A100] bits 0,1,2 (SBL
+             * read-only polls: bit2 poll#1, bit1 poll#2, bit0 gate). */
+            {
+                uint32_t v = 0x00000007;
+                cpu_physical_memory_write(0x073A100, &v, 4);
+            }
+            /* PBL world-switch Thumb-stub (b24, 4B @0x00221EF8):
+             * movs r0,#0; bx lr (return 0 to dispatcher, TZ deferred).
+             * Overwrites ARM trampoline head (recoverable from ELF). */
+            {
+                uint32_t v = 0x47702000;
+                cpu_physical_memory_write(0x00221EF8, &v, 4);
+            }
+             /* S12-stub (b27, 100B @0x08006964, ARM): blx-to-padding
+             * becomes smart dispatcher: ldr ip,[sp,#4] (=pushed lr);
+             * ip==STUBV/even/non-code => ip=poll resumption 0x8007779;
+             * else keep pushed-lr (NO add-sp!); restore r7;
+             * ACK peripheral bit1 ([literal 0x073A100]); bx lr.
+             * PBL would place the real ARM helper here. */
+            {
+                static const uint8_t s12[] = {
+                    0x04, 0xc0, 0x9d, 0xe5, 0x8d, 0x3b, 0x06, 0xe3,
+                    0x00, 0x38, 0x40, 0xe3, 0x03, 0x00, 0x5c, 0xe1,
+                    0x09, 0x00, 0x00, 0x0a, 0x01, 0x00, 0x1c, 0xe3,
+                    0x07, 0x00, 0x00, 0x0a, 0x2c, 0x3c, 0xa0, 0xe1,
+                    0x08, 0x00, 0x53, 0xe3, 0x00, 0x00, 0x00, 0x1a,
+                    0x05, 0x00, 0x00, 0xea, 0x2c, 0x3e, 0xa0, 0xe1,
+                    0x08, 0x00, 0x53, 0xe3, 0x00, 0x00, 0x00, 0x1a,
+                    0x01, 0x00, 0x00, 0xea, 0x79, 0xc7, 0x07, 0xe3,
+                    0x00, 0xc8, 0x40, 0xe3, 0x00, 0x70, 0x0a, 0xe3,
+                    0x73, 0x70, 0x40, 0xe3, 0x0c, 0x30, 0x9f, 0xe5,
+                    0x00, 0x20, 0x93, 0xe5, 0x02, 0x20, 0x82, 0xe3,
+                    0x00, 0x20, 0x83, 0xe5, 0x1e, 0xff, 0x2f, 0xe1,
+                    0x00, 0xa1, 0x73, 0x00,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s12); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s12) - k < 4 ? sizeof(s12) - k : 4;
+                    __builtin_memcpy(&word, &s12[k], n);
+                    cpu_physical_memory_write(0x08006964 + k, &word, 4);
+                }
+            }
+            /* S16-stublet (b27, 64B @0x08006CC0): smart return for the
+             * B66 bx-ip chain. ip==STUBV (fill/maze, never legit) =>
+             * default; ip odd + top 8 => add sp,#8 (pop the helper
+             * frame B66 skipped!) + bx ip (legit pushed-lr, un-leaks);
+             * else default (restore r7, jump poll resumption 0x8007779
+             * via ip). */
+            {
+                static const uint8_t s16[] = {
+                    0xdd, 0xf8, 0x04, 0xc0, 0x46, 0xf6, 0x8d, 0x33,
+                    0xc0, 0xf6, 0x00, 0x03, 0x9c, 0x45, 0x0e, 0xd0,
+                    0x1c, 0xf0, 0x01, 0x0f, 0x0b, 0xd0, 0x5f, 0xea,
+                    0x1c, 0x63, 0x08, 0x2b, 0x01, 0xd1, 0x02, 0xb0,
+                    0x60, 0x47, 0x5f, 0xea, 0x1c, 0x73, 0x08, 0x2b,
+                    0x01, 0xd1, 0x02, 0xb0, 0x60, 0x47, 0x4a, 0xf2,
+                    0x00, 0x07, 0xc0, 0xf2, 0x73, 0x07, 0x47, 0xf2,
+                    0x79, 0x7c, 0xc0, 0xf6, 0x00, 0x0c, 0x60, 0x47,
+                };
+                size_t k;
+                for (k = 0; k < sizeof(s16); k += 4) {
+                    uint32_t word = 0;
+                    size_t n = sizeof(s16) - k < 4 ? sizeof(s16) - k : 4;
+                    __builtin_memcpy(&word, &s16[k], n);
+                    cpu_physical_memory_write(0x08006CC0 + k, &word, 4);
+                }
+            }
+            {
+                uint32_t fill = 0x08006B8D;
                 uint32_t a;
                 for (a = 0x085F0000; a < 0x08600000; a += 4) {
                     cpu_physical_memory_write(a, &fill, 4);
                 }
             }
+            /* IMEM tripwire [0x8600000,0x8600004): drift-sink trap.
+             * Must stay 4B (wider poison corrupts IMEM-stack reads). */
+            {
+                uint32_t p = 0xFFFFFFFF;
+                cpu_physical_memory_write(0x08600000, &p, 4);
+            }
             /* pop-site slots -> stublets (sp values from cpu traces) */
             v = 0x802f65d; /* S1 pop10 @fef0: resume sbl_main */
             cpu_physical_memory_write(0x085FFF14, &v, 4);
@@ -595,6 +929,23 @@ static void saimaa_machine_init(MachineState *machine)
                 printf("saimaa: uefi tables live\n");
             }
             printf("saimaa: slots live\n");
+            /* DDRLOW stub trap overlay: snapshot ELF+stubs, serve on
+             * read/fetch, ignore heap writes (uni parity). Must come
+             * AFTER all stub writes above. Priority 2 > RAMs' 1. */
+            cpu_physical_memory_read(SAIMAA_DSTUB_BASE, saimaa_dstub_mem,
+                                     SAIMAA_DSTUB_SIZE);
+            {
+                MemoryRegion *dstub = g_new(MemoryRegion, 1);
+                memory_region_init_io(dstub, NULL, &saimaa_dstub_ops,
+                                      NULL, "saimaa-dstub",
+                                      SAIMAA_DSTUB_SIZE);
+                memory_region_add_subregion_overlap(sysmem,
+                                                    SAIMAA_DSTUB_BASE,
+                                                    dstub, 2);
+            }
+            /* Freeze stubs: any SBL write here now faults loudly
+             * (data abort with DFAR) instead of silent corruption. */
+            memory_region_set_readonly(saimaa_stubs_mr, true);
         }
         printf("saimaa: SBL1 %s loaded (%zd bytes), entry 0x%lx\n",
                sms->sbl1, sz, (unsigned long)entry);
@@ -604,7 +955,7 @@ static void saimaa_machine_init(MachineState *machine)
 
 static void saimaa_post_reset(void *opaque)
 {
-    uint32_t S = 0x08006900;
+    uint32_t S = 0x08006B80;
     (void)opaque;
     /* Re-apply pokes that live inside ROM-covered ranges (rom_reset
      * restores file bytes over them). */

tools/sbl_uni.py

@@ -0,0 +1,707 @@
+#!/usr/bin/env python3
+"""sbl_uni.py — быстрый harness SBL1 на Unicorn (без пересборки QEMU).
+Карта RAM как в saimaa; ELF-сегменты; UARTDM/TF в stdout; MMIO->0.
+Использование:
+  ./tools/sbl_uni.py fw/img/patch/img0_b17.elf --thumb 0x802f63f [--regs r0=0x220000,...]
+  [--fill 0x085F0000:0x10000:0x08006831] [--poke 0xADDR=0xVAL] [--limit 1000000]
+Печатает: выполнено инстр., последний PC, fault-адрес, UART.
+"""
+import struct
+import sys
+
+from unicorn import (Uc, UC_ARCH_ARM, UC_MODE_THUMB, UC_HOOK_MEM_UNMAPPED,
+                     UC_HOOK_MEM_WRITE, UC_HOOK_CODE)
+from unicorn.arm_const import (UC_ARM_REG_R0, UC_ARM_REG_R1, UC_ARM_REG_R2,
+                                 UC_ARM_REG_R3, UC_ARM_REG_R4, UC_ARM_REG_R5,
+                                 UC_ARM_REG_R6, UC_ARM_REG_R7, UC_ARM_REG_SP,
+                                 UC_ARM_REG_LR, UC_ARM_REG_PC,
+                                 UC_ARM_REG_CPSR)
+_REGS = (UC_ARM_REG_R0, UC_ARM_REG_R1, UC_ARM_REG_R2, UC_ARM_REG_R3,
+         UC_ARM_REG_R4, UC_ARM_REG_R5, UC_ARM_REG_R6, UC_ARM_REG_R7)
+
+RAMS = [
+    (0x00000000, 0x00001000),   # zero page (QEMU-compat: [0] writes succeed)
+    (0x00200000, 0x00100000),   # OCIMEM
+    (0x00400000, 0x07400000),   # RPM range (heap-ish)
+    (0x08000000, 0x00600000),   # DDRLOW
+    (0x08600000, 0x00010000),   # IMEM
+    (0x80000000, 0x07C00000),   # DDRHIGH_A
+    (0x87C00000, 0x00800000),   # SMEM(+SBL BSS)
+    (0x078AF000, 0x00001000),   # UARTDM page
+    (0x13000000, 0x00002000),   # UEFI tables (already above, keep)
+]
+UART_TF = 0x078AF070
+UART_SR = 0x078AF008
+
+
+def load_elf(mu, path):
+    d = open(path, "rb").read()
+    assert d[:4] == b"\x7fELF"
+    entry = struct.unpack("<I", d[24:28])[0]
+    phoff = struct.unpack("<I", d[28:32])[0]
+    phnum = struct.unpack("<H", d[44:46])[0]
+    for k in range(phnum):
+        p = d[phoff + k * 32:phoff + (k + 1) * 32]
+        t, off, v, pa, fsz, msz, fl, al = struct.unpack("<IIIIIIII", p)
+        if t != 1:
+            continue
+        if fsz and v != 0x00220000:  # QEMU parity: skip OCIMEM seg file-bytes
+            mu.mem_write(v, d[off:off + fsz])
+        for z in range(v + fsz, v + msz, 4):
+            mu.mem_write(z, b"\x00\x00\x00\x00")
+    return entry
+
+
+def main():
+    path = sys.argv[1]
+    thumb = int(sys.argv[sys.argv.index("--thumb") + 1], 0)
+    mu = Uc(UC_ARCH_ARM, UC_MODE_THUMB)
+    for base, size in RAMS:
+        mu.mem_map(base, size)
+    uart_out = []
+    unmapped = {"n": 0, "first": []}
+
+    def hook_unmapped(mu, access, addr, size, value, data):
+        unmapped["n"] += 1
+        if len(unmapped["first"]) < 5:
+            unmapped["first"].append((hex(addr), size, access))
+        return False
+
+    mu.hook_add(UC_HOOK_MEM_UNMAPPED, hook_unmapped)
+    # SR reads return TX-ready
+    mu.mem_write(UART_SR, struct.pack("<I", 0xA0))
+    mu.mem_write(0x13000000, struct.pack("<II", 0x49535953, 0x54535953))
+
+    def hook_uw(mu, access, addr, size, value, data):
+        if addr == UART_TF:
+            uart_out.append(chr(value & 0xFF))
+
+    mu.hook_add(UC_HOOK_MEM_WRITE, hook_uw, begin=UART_TF, end=UART_TF)
+    entry = load_elf(mu, path)
+    for r in range(13):
+        mu.reg_write(UC_ARM_REG_R0 + r, 0)
+    if "--regs" in sys.argv:
+        for kv in sys.argv[sys.argv.index("--regs") + 1].split(","):
+            r, v = kv.split("=")
+            mu.reg_write(UC_ARM_REG_R0 + int(r[1:]), int(v, 0))
+    mu.reg_write(UC_ARM_REG_SP, 0x085FFF00)
+    from unicorn.arm_const import UC_ARM_REG_LR as _LR
+    mu.reg_write(_LR, 0x08006B8D)  # QEMU parity: stale lr -> maze
+    mu.reg_write(UC_ARM_REG_CPSR, 0x13)
+    if "--fill" in sys.argv:
+        for spec in sys.argv[sys.argv.index("--fill") + 1].split(","):
+            a, sz, v = spec.split(":")
+            a, sz, v = int(a, 0), int(sz, 0), int(v, 0)
+            mu.mem_write(a, struct.pack("<I", v) * (sz // 4))
+    if "--poke" in sys.argv:
+        for spec in sys.argv[sys.argv.index("--poke") + 1].split(","):
+            a, v = spec.split("=")
+            mu.mem_write(int(a, 0), struct.pack("<I", int(v, 0)))
+    count = [0]
+    last = [0]
+    from collections import deque
+    hist = deque(maxlen=20000)
+    # sled detector: sequential runs (pc == prev+2/4) longer than THRESH
+    # are zero-drifts; log run start (the jump/fall that entered it).
+    _sled_start = [0]
+    _sled_n = [0]
+    _sled_log = []
+    _SLED_TH = 100
+    calls = [] if "--calltrace" in sys.argv else None
+    rets = [] if "--retrace" in sys.argv else None
+    qlog = [] if "--querylog" in sys.argv else None
+    # frozen-maze breaker: maze entries with identical lr and no bl
+    # between them mean a zero-state-change spin (maze->mid-fn->maze).
+    # After K, redirect once to poll resumption (fresh bls unfreeze lr).
+    _fmlr = [None]
+    _fmct = [0]
+    _FMK = 50
+    _FMPC = 0x8007779
+    _traps = []
+    _trap_n = [0]
+    if "--trap" in sys.argv:
+        _traps = [int(x, 0) for x in sys.argv[sys.argv.index("--trap") + 1].split(",")]
+    if "--trapn" in sys.argv:
+        _trap_n = [int(sys.argv[sys.argv.index("--trapn") + 1])]
+    _spmax = [0x085FFF00]
+    _splog = [] if "--spytrack" in sys.argv else None
+    elog = [] if "--entrylog" in sys.argv else None
+    _prev = [0]
+    _eranges = []
+    _inrng = []
+    if "--entrylog" in sys.argv:
+        for _spec in sys.argv[sys.argv.index("--entrylog") + 1].split(","):
+            _lo, _sz = _spec.split(":")
+            _lo, _sz = int(_lo, 0), int(_sz, 0)
+            _eranges.append((_lo, _lo + _sz))
+            _inrng.append(False)
+    from unicorn.arm_const import UC_ARM_REG_CPSR as _CPSR
+
+    def hook_code(mu, addr, size, data):
+        count[0] += 1
+        last[0] = addr
+        _pp = _prev[0]
+        _prev[0] = addr
+        if addr == _pp + 2 or addr == _pp + 4:
+            if _sled_n[0] == 0:
+                _sled_start[0] = _pp
+            _sled_n[0] += 1
+        else:
+            if _sled_n[0] >= _SLED_TH and len(_sled_log) < 20:
+                _sled_log.append((count[0], _sled_start[0], _pp, _sled_n[0]))
+            _sled_n[0] = 0
+        if 0x08006B8C <= addr < 0x08006B9E:
+            try:
+                _lr = mu.reg_read(_LR)
+            except Exception:
+                _lr = None
+            if _lr == _fmlr[0]:
+                _fmct[0] += 1
+                if _fmct[0] >= _FMK:
+                    _fmct[0] = 0
+                    _fmlr[0] = None
+                    mu.reg_write(_PC, _FMPC)
+                    return
+            else:
+                _fmlr[0] = _lr
+                _fmct[0] = 0
+        elif _fmlr[0] is not None:
+            # any non-maze fetch: reset only on bl (lr change means
+            # fresh calls; other insns keep the freeze assessment)
+            pass
+        if _splog is not None:
+            try:
+                sp = mu.reg_read(UC_ARM_REG_SP)
+            except Exception:
+                sp = 0
+            if sp > _spmax[0] and len(_splog) < 40:
+                _spmax[0] = sp
+                _splog.append((count[0], addr, sp))
+        if _dirty[0]:
+            # deferred restore: MEM_WRITE hooks fire BEFORE the write lands,
+            # so in-hook restore gets overwritten by the instruction itself
+            _inww[0] = True
+            try:
+                for a, b in _dirty[1]:
+                    mu.mem_write(a, stub_snap[a - STUB_LO:b - STUB_LO])
+            finally:
+                _inww[0] = False
+                _dirty[0] = False
+                del _dirty[1][:]
+        try:
+            thumb = (mu.reg_read(_CPSR) >> 5) & 1
+        except Exception:
+            thumb = 1
+        hist.append((addr, thumb))
+        if _traps and addr in _traps:
+            try:
+                vals = [mu.reg_read(_REGS[i]) for i in range(8)]
+            except Exception:
+                vals = []
+            print(f"TRAP {addr:#x} insn#{count[0]} r0-r7={[hex(x) for x in vals]} sp={mu.reg_read(UC_ARM_REG_SP):#x} lr={mu.reg_read(_LR):#x}")
+            try:
+                print(f"  [73A100]={bytes(mu.mem_read(0x073A100, 4)).hex()}")
+            except Exception as e:
+                print(f"  [73A100] unreadable: {e}")
+            try:
+                from capstone import Cs, CS_ARCH_ARM, CS_MODE_THUMB, CS_MODE_ARM
+                md_t = Cs(CS_ARCH_ARM, CS_MODE_THUMB)
+                md_a = Cs(CS_ARCH_ARM, CS_MODE_ARM)
+                for _pc, _th in list(hist)[-60:]:
+                    try:
+                        _b = _pc & ~1
+                        _c = mu.mem_read(_b, 8 if _th else 12)
+                        for _ins in (md_t if _th else md_a).disasm(_c, _b):
+                            print(f"  hist {_b:#x}{'t' if _th else 'a'}: {_ins.mnemonic} {_ins.op_str}")
+                            break
+                    except Exception:
+                        print(f"  hist {_pc:#x}: <unreadable>")
+            except ImportError:
+                pass
+            _trap_n[0] -= 1
+            if _trap_n[0] <= 0 and "--trapn" in sys.argv:
+                mu.emu_stop()
+                return
+            if "--trapn" not in sys.argv:
+                mu.emu_stop()
+                return
+        if qlog is not None and addr in (0x8017EA4, 0x8017C60):
+            try:
+                vals = [mu.reg_read(_REGS[i]) for i in range(8)]
+            except Exception:
+                vals = []
+            qlog.append((addr, vals,
+                         mu.reg_read(UC_ARM_REG_SP),
+                         mu.reg_read(UC_ARM_REG_LR)))
+        if elog is not None:
+            for _i, (_elo, _ehi) in enumerate(_eranges):
+                inside = _elo <= addr < _ehi
+                if inside and not _inrng[_i]:
+                    if len(elog) < 20000:
+                        try:
+                            _th = (mu.reg_read(_CPSR) >> 5) & 1
+                        except Exception:
+                            _th = -1
+                        elog.append((count[0], _prev[0], addr, _th))
+                if not inside and _inrng[_i]:
+                    if len(elog) < 20000:
+                        elog.append((count[0], _prev[0], -addr, 2))
+                _inrng[_i] = inside
+        if calls is not None and size in (2, 4):
+            try:
+                if size == 2:
+                    h = struct.unpack("<H", mu.mem_read(addr & ~1, 2))[0]
+                    is_blx = (h & 0xFF80) == 0x4780
+                    rm = (h >> 3) & 7 if is_blx else -1
+                    # bx Rm (jump, no link): 010001110 Rm:000
+                    is_bx = (h & 0xFF80) == 0x4700
+                    if is_bx:
+                        bxrm = (h >> 3) & 15
+                        try:
+                            if bxrm < 8:
+                                bxt = mu.reg_read(_REGS[bxrm])
+                            elif bxrm == 14:
+                                bxt = mu.reg_read(_LR)
+                            else:
+                                bxt = -1
+                        except Exception:
+                            bxt = -1
+                        if len(calls) < 8000:
+                            calls.append((addr, 100 + bxrm, bxt))
+                    # pop {..,pc} (0xBD00+): log stacked target
+                    if calls is not None and (h & 0xFF00) == 0xBD00:
+                        try:
+                            _sp = mu.reg_read(UC_ARM_REG_SP)
+                            _tgt = struct.unpack("<I", mu.mem_read(_sp, 4))[0]
+                        except Exception:
+                            _tgt = -1
+                        if len(calls) < 12000:
+                            calls.append((addr, 300, _tgt))
+                    if rets is not None and 0x8017E00 <= addr < 0x8018000:
+                        # pop {..,pc} (0xBD00+) or bx lr (0x4770)
+                        if h == 0x4770 or (h & 0xFF00) == 0xBD00:
+                            try:
+                                rv = mu.reg_read(_REGS[0])
+                            except Exception:
+                                rv = -1
+                            if len(rets) < 200:
+                                rets.append((addr, rv))
+                else:
+                    w = struct.unpack("<I", mu.mem_read(addr & ~3, 4))[0]
+                    # ARM bx/blx Rm: cond(4) 0001001011111111111100x1 Rm(4)
+                    is_blx = (w & 0x0FFFFFF0) in (0x012FFF10, 0x012FFF30)
+                    rm = w & 15 if is_blx else -1
+                    tgt = -1
+                    if not is_blx and calls is not None:
+                        # Thumb32 BL: h0=11110S+imm10, h1=1111x+J+imm11
+                        h0 = w & 0xFFFF
+                        h1 = (w >> 16) & 0xFFFF
+                        if (h0 & 0xF800) == 0xF000 and (h1 & 0xE000) == 0xE000:
+                            S = (h0 >> 10) & 1
+                            J1 = (h1 >> 13) & 1
+                            J2 = (h1 >> 11) & 1
+                            I1 = 1 - (J1 ^ S)
+                            I2 = 1 - (J2 ^ S)
+                            off = (I1 << 23) | (I2 << 22) | ((h0 & 0x3FF) << 12) | ((h1 & 0x7FF) << 1)
+                            if S:
+                                off -= 0x1000000
+                            if len(calls) < 8000:
+                                calls.append((addr, 200, (addr + 4 + off) & 0xFFFFFFFF))
+                    if rets is not None and 0x8017E00 <= addr < 0x8018000:
+                        # pop.w {...,pc}: 0xE8BDxxxx with bit15 set
+                        if (w & 0xFFFF8000) == 0xE8BD8000:
+                            try:
+                                rv = mu.reg_read(_REGS[0])
+                            except Exception:
+                                rv = -1
+                            if len(rets) < 200:
+                                rets.append((addr, rv))
+                if is_blx:
+                    try:
+                        tgt = mu.reg_read(_REGS[rm])
+                    except Exception:
+                        tgt = -1
+                    if len(calls) < 4000:
+                        calls.append((addr, rm, tgt))
+            except Exception:
+                return
+
+    mu.hook_add(UC_HOOK_CODE, hook_code)
+    # null-fetch redirect: branch-to-0 lands in maze (movs r0,#0; bx lr),
+    # immune to [0]-writes that clobber a static maze copy
+    from unicorn.arm_const import UC_ARM_REG_PC as _PC
+    _nulllog = []
+    _halt_on_null = "--haltonnull" in sys.argv
+
+    def hook_null(mu, addr, size, data):
+        if len(_nulllog) < 10:
+            try:
+                r = [mu.reg_read(_REGS[i]) for i in range(4)]
+                lr = mu.reg_read(_LR)
+                sp = mu.reg_read(UC_ARM_REG_SP)
+            except Exception:
+                r, lr, sp = [], -1, -1
+            # walk back: find the branch source from hist
+            _nulllog.append((count[0], r, lr, sp))
+        if _halt_on_null and len(_nulllog) == 1:
+            return False
+        mu.reg_write(_PC, STUB + 1)
+
+    mu.hook_add(UC_HOOK_CODE, hook_null, begin=0x0, end=0xFFF)
+    if "--watchcode" in sys.argv:
+        from unicorn.arm_const import UC_ARM_REG_PC as _PC2
+        wlog = []
+
+        def hook_ww(mu, access, addr, size, value, data):
+            if len(wlog) < 30:
+                try:
+                    pc = mu.reg_read(_PC2)
+                except Exception:
+                    pc = -1
+                wlog.append((pc, addr, size, value))
+
+        mu.hook_add(UC_HOOK_MEM_WRITE, hook_ww,
+                    begin=0x8005800, end=0x8039254)
+        import atexit
+
+        def _dump():
+            print(f"code-writes: {len(wlog)}")
+            for pc, a, sz, v in wlog[:30]:
+                print(f"  pc={pc:#x} [{a:#x}] sz={sz} val={v:#x}")
+        atexit.register(_dump)
+    # --- machine scaffold parity (saimaa.c values, verbatim) ---
+    def W32(a, v):
+        mu.mem_write(a, struct.pack("<I", v & 0xFFFFFFFF))
+    S, STUB, STUBV = 0x08006B80, 0x08006B8C, 0x08006B8D
+    W32(S, S + 8)
+    W32(S + 4, STUBV)
+    W32(S + 8, STUBV)
+    # maze blob (24B @STUB): movs r0,#0; return 0 via lr ONLY if lr is
+    # odd Thumb in DDR (top 8); even lr (stale mid-fn raw addr!) would
+    # ARM-drift (STM-clobbers + IMEM fault), so hang loud instead.
+    # Overlaps S6 head ([B9E,BA4), S6 dormant, documented).
+    mu.mem_write(STUB, bytes([
+        0x00, 0x20, 0x4f, 0xea, 0x1e, 0x6c, 0xbc, 0xf1,
+        0x08, 0x0f, 0x03, 0xd1, 0x1e, 0xf0, 0x01, 0x0f,
+        0x00, 0xd0, 0x70, 0x47, 0x72, 0xb6, 0xfd, 0xe7]))
+    print("maze-rb:", bytes(mu.mem_read(STUB, 24)).hex())
+    for a in (0x087c29d18, 0x08050b28, 0x08070000):
+        W32(a, S)
+    # S6 stublet DISABLED (maze v2 overlaps its head; S6 dormant,
+    # never entered (b14 goes to 0x8005B9E, not here). Bytes kept for
+    # reference (would live at 0x08006B9E):
+    # mu.mem_write(0x08006B9E, bytes([
+    #     0xbd, 0xe8, 0xfc, 0x47, 0x43, 0xf6, 0xa5, 0x3e,
+    #     0xc0, 0xf6, 0x01, 0x0e, 0x70, 0x47]))
+    # banner stub (34B @0x08006BAC, prints r0 via UARTDM)
+    mu.mem_write(0x08006BAC, bytes([
+        0x10, 0xb5, 0x4f, 0xf2, 0x00, 0x03, 0xc0, 0xf2,
+        0x8a, 0x73, 0x21, 0x21, 0x19, 0x67, 0x10, 0xf8,
+        0x01, 0x1b, 0x09, 0xb1, 0x19, 0x67, 0xfa, 0xe7,
+        0x4f, 0xf2, 0x94, 0x6e, 0xc0, 0xf6, 0x02, 0x0e,
+        0x70, 0x47]))
+    # PBL stub (52B @0x080068C0)
+    mu.mem_write(0x080068C0, bytes([
+        0xf0, 0xb5, 0x46, 0xf6, 0x81, 0x3c, 0xc0, 0xf6,
+        0x00, 0x0c, 0xe0, 0x47, 0x40, 0xf2, 0x00, 0x00,
+        0xc0, 0xf2, 0x22, 0x00, 0x05, 0x46, 0x40, 0xf2,
+        0x00, 0x0a, 0xc0, 0xf6, 0x06, 0x0a, 0x41, 0xf2,
+        0x00, 0x0b, 0xc0, 0xf6, 0x06, 0x0b, 0x4f, 0xf2,
+        0x3f, 0x6e, 0xc0, 0xf6, 0x02, 0x0e, 0x70, 0x47,
+        0x00, 0x00, 0x00, 0x00]))
+    # S7 stublet (52B @0x08006C10, b18): delay-epilogue exit router.
+    # cmp r4,#0xff: iterate (r4<=0xff -> restart 0x8013ba5) or done
+    # (restore r4/r5/sl/fp, return to sbl_main 0x802f681).
+    mu.mem_write(0x08006C10, bytes.fromhex(
+        "ff2c13d940f20004c0f2220440f20005c0f2220540f2000a"
+        "c0f6060a41f2000bc0f6060bdff800f081f60208dff800f0"
+        "a53b0108"))
+    # S9 stublet (62B @0x08006C48, b22): labyrinth-epilogue exit router.
+    # ldr r3,[sp,#0x14]: nonzero -> original pop.w (legit return);
+    # zero (empty scaffold frames) -> restore r0/r4/r5/sl/fp, reset sp
+    # to 0x085FFF00 (fresh STUBV-fill stack), jump to sbl_main 0x802f69d
+    # via ip (no literal loads).
+    mu.mem_write(0x08006C48, bytes.fromhex(
+        "059b0bb1bde8f081002040f20004c0f2220440f20005c0f2"
+        "220540f2000ac0f6060a41f2000bc0f6060b4ff60073c0f6"
+        "5f039d464ff29d6cc0f6020c6047"))
+    # S10 stublet (10B @0x08006C88, b23): rollback-exit router.
+    # movw/movt ip + bx ip to sbl_main 0x802f6a3.
+    mu.mem_write(0x08006C88, bytes.fromhex("4ff2a36cc0f6020c6047"))
+    # S18 stublet (10B @0x08006D22, b29): poll-epilogue router.
+    # movw/movt ip + bx ip to sbl_main 0x802f6a9 (post-dispatcher).
+    mu.mem_write(0x08006D22, bytes.fromhex("4ff2a96cc0f6020c6047"))
+    # S21 stublet (32B @0x08006D5A, b38): table-walker router.
+    # [sp+0x24] valid-code? => original pop.w (legit return);
+    # else b.w S19 (sbl_main post-logger restart, bounded table work).
+    mu.mem_write(0x08006D5A, bytes.fromhex(
+        "099b13f0010f09d01a0e082a01d1bde8f09f1a0f082a"
+        "01d1bde8f09ffff7d9bf"))
+    # S20 stublet (34B @0x08006D38, b37): clock-dispatcher router.
+    # [sp+0x2C] valid-code? => original pop.w (legit return);
+    # else r0=sl (PBL tables) + b.w dispatcher re-run (idempotent math).
+    mu.mem_write(0x08006D38, bytes.fromhex(
+        "0b9b13f0010f09d01a0e082a01d1bde8fe8f1a0f082a"
+        "01d1bde8fe8f504607f013bf"))
+    # S17 safe-logger (50B @0x08006CF0, b26): print r1-string via
+    # UARTDM-TF only if r1 in rodata [0x804B800,0x8058000); pop-return.
+    mu.mem_write(0x08006CF0, bytes.fromhex(
+        "4bf60003c0f60403994211d348f20003c0f605039942"
+        "0bd210b54ff20003c0f28a730c4614f8010b08b11867fae710bd7047"))
+    # PBL version-table fabrication (b23): [0x080528D0] = STUBV so the
+    # rollback query's blx calls maze (returns 0 = versions OK).
+    W32(0x080528D0, STUBV)
+    # PBL dispatcher-table fabrication (b33): [0x80164F0] (literal pool,
+    # image default points at hash blob!) retargeted to scratch struct
+    # at 0x08024000 (OCIMEM, out of stack-march paths!) with [0x18]=2
+    # (ready-flag -> world-switch dispatch; TZ deferred via OCIMEM stub).
+    W32(0x08024000, 0x08024000)
+    W32(0x08024014, 0x00000002)
+    W32(0x080164F0, 0x08024000)
+    # HW-ready bit fabrication (b25): [0x073A100] bits 0,1,2 = peripheral
+    # status done-bits (poll#1 checks bit2 (lsls#29), poll#2 bit1 (lsls#30),
+    # post-poll gate bit0 (lsls#31)). SBL only reads it, never writes.
+    W32(0x073A100, 0x00000007)
+    # PBL world-switch Thumb-stub (b24): the ARM trampoline at 0x221EF8
+    # is entered in Thumb (mode confusion at dispatch); real ARM would
+    # save regs/CPS to MON and call TZ (not loaded yet). Stub returns 0
+    # to the dispatcher via stale lr (TZ bringup deferred).
+    # NOTE: overwrites 4B of real ARM trampoline (recoverable from ELF).
+    mu.mem_write(0x00221EF8, bytes([0x00, 0x20, 0x70, 0x47]))
+    # S12-stub (b27, 100B @0x08006964, ARM): blx-to-padding becomes
+    # smart dispatcher: ldr ip,[sp,#4] (=pushed lr); ip==STUBV/even/
+    # non-code => ip=poll resumption 0x8007779; else keep pushed-lr
+    # (NO add-sp: marches on mid-entries! un-leak dropped, helper-leak
+    # down is harmless); restore r7; ACK peripheral bit1; bx lr.
+    # PBL would place the real ARM helper here.
+    mu.mem_write(0x08006964, bytes.fromhex(
+        "04c09de58d3b06e3003840e303005ce10900000a0100"
+        "1ce30700000a2c3ca0e1080053e30000001a050000ea"
+        "2c3ea0e1080053e30000001a010000ea79c707e300c8"
+        "40e300700ae3737040e30c309fe5002093e5022082e3"
+        "002083e51eff2fe100a17300"))
+    # S16-stublet (b27, 64B @0x08006CC0): smart return for B66 bx ip.
+    # ip==STUBV (fill/maze, never a legit return) => default;
+    # ip odd + top 8 (code) => add sp,#8 (pop the helper frame B66
+    # skipped!) + bx ip (legit pushed-lr return, un-leaks the stack);
+    # else default (restore r7, jump poll resumption 0x8007779).
+    mu.mem_write(0x08006CC0, bytes.fromhex(
+        "ddf804c046f68d33c0f600039c450ed01cf0010f0bd0"
+        "5fea1c63082b01d102b060475fea1c73082b01d102b0"
+        "60474af20007c0f2730747f2797cc0f6000c6047"))
+    # stack zone fill (STUBV odd)
+    mu.mem_write(0x085F0000, struct.pack("<I", STUBV) * (0x10000 // 4))
+    # IMEM tripwire [0x8600000,0x8600004): drift-sink trap (UDF).
+    # Fill-drift falls through here sequentially; faults LOUDLY at
+    # entry. Must stay 4B: wider poison corrupts IMEM-stack reads
+    # (SBL frames live at 0x860000C+, counters underflow on 0xFF!).
+    mu.mem_write(0x08600000, b"\xff\xff\xff\xff")
+    # pop-site slots (sp values from traces)
+    for a, v in [(0x085FFF14, 0x802f65d), (0x085FFF1C, 0x802f665),
+                 (0x085FFF44, 0x8013ba5), (0x085FFF5C, 0x802f681),
+                 (0x085FFF74, 0x8016183)]:
+        W32(a, v)
+    W32(0x085FFF58, 0)
+    W32(0x085FFF34, 0)
+    W32(0x085FFF54, 0)
+    W32(0x085FFF6C, 0)
+    W32(0x08600944, 0)  # IMEM poll magic
+    # minimal UEFI tables (ST/BS/RS/CO + ret0 stubs)
+    ST, BS, RS, CO, STUBS, LI, HEAP = (0x13000000, 0x13000100, 0x13000200,
+                                       0x13000300, 0x13000400, 0x13000500,
+                                       0x13001000)
+    for i in range(24):
+        mu.mem_write(STUBS + i * 8, bytes([0x00, 0x20, 0x70, 0x47,
+                                           0x00, 0x00, 0x00, 0x00]))
+        W32(BS + 24 + i * 4, STUBS + i * 8 + 1)
+    W32(BS, 0x56524553)
+    W32(ST, 0x49535953)
+    W32(ST + 4, 0x54535953)
+    W32(ST + 12, 0x00020000)
+    W32(ST + 16, 0x13000048)
+    W32(ST + 40, CO)
+    W32(ST + 44, CO)
+    W32(ST + 56, RS)
+    W32(ST + 60, BS)
+    W32(LI, 0x3130414d)
+    W32(HEAP - 4, HEAP)
+    # NOTE: no vendor string bytes (QEMU leaves zeros at 0x13000048)
+    # map UEFI table region (already in RAMS? add 0x13000000)
+    # --- end parity ---
+    # stub range is readonly on HW/QEMU: ignore SBL heap-clobber writes,
+    # keep maze/stublets intact (restore-on-write with reentrancy guard)
+    STUB_LO, STUB_HI = 0x080068C0, 0x08006D80  # +S7/S9/S10/S16/S17/S18/S20/S21
+    stub_snap = bytes(mu.mem_read(STUB_LO, STUB_HI - STUB_LO))
+    _inww = [False]
+    _swn = [0]
+    _swlog = []
+    _dirty = [False, []]
+    print("stub-snap:", stub_snap[:24].hex())
+    print("stub-snap-maze:", stub_snap[0x2CC:0x2CC + 18].hex())
+
+    def hook_stubw(mu, access, addr, size, value, data):
+        if _inww[0]:
+            return
+        if addr + size <= STUB_LO or addr >= STUB_HI:
+            return
+        _swn[0] += 1
+        if len(_swlog) < 8:
+            try:
+                _pc = mu.reg_read(_PC)
+            except Exception:
+                _pc = -1
+            _swlog.append((_pc, addr, size))
+        # record for deferred restore (see hook_code); in-hook restore
+        # would be overwritten by the faulting instruction itself
+        a = max(addr, STUB_LO)
+        b = min(addr + size, STUB_HI)
+        if b > a and len(_dirty[1]) < 16:
+            _dirty[1].append((a, b))
+            _dirty[0] = True
+
+    #hook all DDRLOW, filter in python (catches writes starting below LO)
+    mu.hook_add(UC_HOOK_MEM_WRITE, hook_stubw,
+                begin=0x08000000, end=0x085FFFFF)
+    if "--watchmaze" in sys.argv:
+        from unicorn.arm_const import UC_ARM_REG_PC as _PC3
+        mzlog = []
+
+        def hook_mz(mu, access, addr, size, value, data):
+            if len(mzlog) < 40:
+                try:
+                    pc = mu.reg_read(_PC3)
+                except Exception:
+                    pc = -1
+                mzlog.append((pc, addr, size, value))
+
+        mu.hook_add(UC_HOOK_MEM_WRITE, hook_mz,
+                    begin=0x08006B8C, end=0x08006B9D)
+        import atexit as _ax
+
+        def _dmz():
+            print(f"maze-writes: {len(mzlog)}")
+            for pc, a, sz, v in mzlog[:40]:
+                print(f"  pc={pc:#x} [{a:#x}] sz={sz} val={v:#x}")
+        _ax.register(_dmz)
+    if "--watchlow" in sys.argv:
+        _spec = sys.argv[sys.argv.index("--watchlow") + 1].split(":")
+        _wlo, _whi = int(_spec[0], 0), int(_spec[0], 0) + int(_spec[1], 0)
+        from unicorn.arm_const import UC_ARM_REG_PC as _PC4
+        wlolog = []
+
+        def hook_wlo(mu, access, addr, size, value, data):
+            if len(wlolog) < 60:
+                try:
+                    pc = mu.reg_read(_PC4)
+                except Exception:
+                    pc = -1
+                wlolog.append((pc, addr, size, value))
+
+        mu.hook_add(UC_HOOK_MEM_WRITE, hook_wlo,
+                    begin=0x00200000, end=0x085FFFFF)
+        import atexit as _ax2
+
+        def _dwlo():
+            inn = [(pc, a, sz, v) for pc, a, sz, v in wlolog
+                   if not (a + sz <= _wlo or a >= _whi)]
+            print(f"low-writes-in-range: {len(inn)}")
+            for pc, a, sz, v in inn[:60]:
+                print(f"  pc={pc:#x} [{a:#x}] sz={sz} val={v:#x}")
+        _ax2.register(_dwlo)
+    limit = 2000000
+    if "--limit" in sys.argv:
+        limit = int(sys.argv[sys.argv.index("--limit") + 1])
+    try:
+        mu.emu_start(thumb | 1, 0xFFFFFFFF, 0, limit)
+    except Exception as e:
+        print("EMU-EXC:", str(e)[:120])
+    print("insns:", count[0], "last-pc:", hex(last[0]))
+    print("unmapped:", unmapped["n"], unmapped["first"][:4])
+    print("regs:",
+          " ".join(f"r{i}={mu.reg_read(_REGS[i]):#x}" for i in range(8)),
+          f"sp={mu.reg_read(UC_ARM_REG_SP):#x}",
+          f"lr={mu.reg_read(UC_ARM_REG_LR):#x}",
+          f"pc={mu.reg_read(UC_ARM_REG_PC):#x}")
+    try:
+        from capstone import Cs, CS_ARCH_ARM, CS_MODE_THUMB, CS_MODE_ARM
+        md_t = Cs(CS_ARCH_ARM, CS_MODE_THUMB)
+        md_a = Cs(CS_ARCH_ARM, CS_MODE_ARM)
+        real = [(pc, th) for pc, th in hist
+                if not (0 <= (pc & ~1) < 0x1000)][-48:]
+        for pc, thumb in real:
+            try:
+                base = pc & ~1
+                code = mu.mem_read(base, 8 if thumb else 12)
+                md = md_t if thumb else md_a
+                for ins in md.disasm(code, base):
+                    print(f"  {base:#x}{'t' if thumb else 'a'}: {ins.mnemonic} {ins.op_str}")
+                    break
+            except Exception:
+                print(f"  {pc:#x}: <unreadable>")
+    except ImportError:
+        print("hist:", [hex(pc) for pc, _ in list(hist)[-12:]])
+    if _sled_log or _sled_n[0] >= _SLED_TH:
+        if _sled_n[0] >= _SLED_TH:
+            _sled_log.append((count[0], _sled_start[0], last[0], _sled_n[0]))
+        print(f"sleds: {len(_sled_log)}")
+        for n, s, e, c in _sled_log:
+            print(f"  insn#{n} [{s:#x}..{e:#x}] len={c}")
+    if elog is not None:
+        print(f"entries: {len(elog)}")
+        for n, prev, cur, th in elog[:10]:
+            if cur < 0:
+                print(f"  FIRST EXIT insn#{n} {prev:#x} -> {(-cur):#x}")
+            else:
+                print(f"  FIRST insn#{n} {prev:#x} -> {cur:#x}{'t' if th == 1 else ('a' if th == 0 else '?')}")
+        for n, prev, cur, th in elog[-10:]:
+            if cur < 0:
+                print(f"  LAST EXIT insn#{n} {prev:#x} -> {(-cur):#x}")
+            else:
+                print(f"  LAST insn#{n} {prev:#x} -> {cur:#x}{'t' if th == 1 else ('a' if th == 0 else '?')}")
+    if _splog is not None:
+        print(f"sp-highs: {len(_splog)}")
+        for n, pc, sp in _splog:
+            print(f"  insn#{n} {pc:#x} sp={sp:#x}")
+    if _nulllog:
+        print(f"null-entries: {len(_nulllog)}")
+        for n, r, lr, sp in _nulllog:
+            print(f"  insn#{n} r0-r3={[hex(x) for x in r]} lr={lr:#x} sp={sp:#x}")
+    if qlog is not None:
+        print(f"querylog: {len(qlog)}")
+        for pc, r, sp, lr in qlog[:20]:
+            print(f"  {pc:#x} r0-r7={[hex(x) for x in r]} sp={sp:#x} lr={lr:#x}")
+    if rets is not None:
+        print(f"rets-in-range: {len(rets)}")
+        for pc, rv in rets[-20:]:
+            print(f"  ret {pc:#x} r0={rv:#x}" if rv >= 0 else f"  ret {pc:#x} r0=?")
+    if calls is not None:
+        print(f"blx-calls: {len(calls)}")
+        bls = [(pc, tgt) for pc, rm, tgt in calls if rm == 200]
+        print(f"bl-immed: {len(bls)}")
+        for pc, tgt in bls[-30:]:
+            print(f"  {pc:#x}: bl #{tgt:#x}")
+        pops = [(pc, tgt) for pc, rm, tgt in calls if rm == 300]
+        print(f"pop-pc: {len(pops)}")
+        for pc, tgt in pops[:15]:
+            print(f"  FIRST {pc:#x}: pop-> {tgt:#x}" if tgt >= 0 else f"  FIRST {pc:#x}: pop-> ?")
+        for pc, tgt in pops[-15:]:
+            print(f"  LAST {pc:#x}: pop-> {tgt:#x}" if tgt >= 0 else f"  LAST {pc:#x}: pop-> ?")
+        for pc, rm, tgt in calls[-12:]:
+            if rm == 200:
+                continue
+            tag = f"bx r{rm - 100}" if rm >= 100 else f"blx r{rm}"
+            print(f"  {pc:#x}: {tag} -> {tgt:#x}" if tgt >= 0 else f"  {pc:#x}: {tag} -> ?")
+    if "--dumpmem" in sys.argv:
+        for spec in sys.argv[sys.argv.index("--dumpmem") + 1].split(","):
+            a, sz = spec.split(":")
+            a, sz = int(a, 0), int(sz, 0)
+            print(f"mem[{a:#x}+{sz:#x}]:")
+            print(bytes(mu.mem_read(a, sz)).hex())
+    print("UART:", "".join(uart_out)[:160])
+    print(f"stubguard-hits: {_swn[0]}")
+    for _pc, _a, _s in _swlog:
+        print(f"  guarded pc={_pc:#x} [{_a:#x}] sz={_s}")
+
+
+if __name__ == "__main__":
+    main()