b40: S12-wrapper pop restore kills 8B/call leak; uni DDRHI parity; run-sbl b40

docs/sbl-bringup.md

@@ -229,3 +229,15 @@
   UARTDM-страница) замаплен в uni как leak-sink (+8M runway).
   QEMU не тронут (в гэпе периферия; короткие прогоны не достают).
   UART все еще пуст.
+
+## b40: S12-wrapper pop restore (2026-09-14)
+- 105M-фолт раскрыт ловушкой записи в текст: push.w на 0x80078DC писал
+  по sp=0x802F688 (стек в тексте sbl_main), далее fetch по затертым
+  байтам. Корень: b25 заменил pop r4,pc на bx ip во враппере S12
+  (0x8007B66), каждый вызов сиротил 8B фрейма (около 58B/Kinsn);
+  к 105M sp съехал, pop.w с восстановлением sp/pc (0x800EB94) взял
+  чужой stacked-lr. b40 вернул pop (2B), r0=0 снова по контракту.
+- Побочка: ранний clock-путь теперь идет честно (раньше диверт его
+  пропускал): IMEM-стек, чтение за верхом IMEM. Uni не мапил DDRHI,
+  QEMU мапит: добавлен паритет (0x08610000, 16M-64K). b40 на 2M чист,
+  sp живет в DDRHI, unmapped 0. Дефолт run-sbl теперь b40.

qemu/run-sbl.sh

@@ -5,7 +5,7 @@
 set -e
 cd "$(dirname "$0")/.."
 QEMU=~/qemu-src/qemu-8.2.2/build/qemu-system-arm
-B=${1:-b39}
+B=${1:-b40}
 IMG=fw/img/emmc-real.img
 if [ ! -f "$IMG" ]; then
   echo "no $IMG — сначала: python3 tools/ffu_extract.py fw/dl/RM1127_059X5R5_orig.ffu --emmc-user fw/img/user-real.bin"

tools/sbl_patch.py

@@ -2553,8 +2553,148 @@ def main():
 
     d[o:o + 2] = bytes.fromhex('c046')
 
+    
     save(d, 'img0_b39.elf')
 
+    # b40 = b39 + S12-wrapper pop restore (0x8007B66 bx ip -> pop):
+    # b25 chained the wrapper pop to bx ip; with S12 preserving sp the
+    # frame is intact, while bx ip orphaned 8B per call (105M sp-smash).
+    # Content below = b39 body with the 0x8007B66 hunk reverted to pop.
+    # naked clock leaf never sets r3 (stale stacked return addr,
+    # observed 0x8007975); the store writes 0xffff over live code,
+    # faulting the next fetch there. Peripheral write, nothing in emu
+    # reads it back (polls use scaffolded [0x73A100]). Same rationale
+    # as the b17 stm NOP. 2B->2B, no eating.
+    d = bytearray(base)
+
+    for va, (a, b) in BLX2BL.items():
+
+        set4(d, va, b)
+
+    for va, exp, new in [
+
+        (0x802F664, 'e4f77efc', 'c046c046'),
+
+        (0x802F672, 'e6f7bbfa', 'c046c046'),
+
+        (0x802F678, 'e9f7fef9', 'c046c046'),
+
+        (0x801B71E, '9847fee7', 'c046c046'),
+
+        (0x8008396, 'fef780f9', 'c046c046'),
+
+        (0x8016184, 'fdf790fc', 'c046c046'),
+
+        (0x801B76C, '280b0508', '00000708'),
+
+        (0x8017AFC, 'bde8fc87', 'eef74fb8'),
+
+        (0x802F690, '9847fee7', 'c046c046'),
+
+        (0x801671C, '9847fee7', 'c046c046'),
+
+        (0x8013BAC, 'f8bd0a4a', 'f3f730b8'),
+
+        (0x801B6BE, 'bde8f081', 'ebf7c3ba'),
+
+        (0x8016720, '10bd0000', 'f0f7b2ba'),
+
+        (0x8034810, 'bde8fc87', 'd2f76eba'),
+
+        (0x8007798, 'bde8f08f', 'fff7c3ba'),
+
+        (0x8016334, 'f0f7a4ff', 'c046c046'),
+
+        (0x8016342, 'f0f79eef', 'f0f79dbf'),
+
+        (0x802F9A0, 'ffe70648', 'd7f7c4b9'),
+
+        (0x802F9B2, 'fee78817', 'd7f7bbb9'),
+
+        (0x802F9A2, 'c4b943f2', 'd7f7c3b9'),
+
+        (0x800EB96, 'bde8fe8f', 'f8f7cfb8'),
+
+        (0x8007916, 'bde8f09f', 'fff720ba'),
+
+    ]:
+
+        assert hx(d, va) == exp, (hex(va), hx(d, va))
+
+        set4(d, va, new)
+
+    # Full bl idiom at 0x8006682 (bl + pop, 6B) -> NOPs (see b18).
+    o = off(0x8006682)
+
+    assert d[o:o + 6].hex() == '00f001f807bd', hx(d, 0x8006682, 6)
+
+    d[o:o + 6] = bytes.fromhex('c046c046c046')
+
+    o = off(0x8013AA8)
+
+    assert d[o:o + 4].hex() == '07f026be', hx(d, 0x8013AA8)
+
+    d[o:o + 4] = bytes.fromhex('c046c046')
+
+    o = off(0x8013BA6)
+
+    assert d[o:o + 2].hex() == 'b888', (hex(0x8013BA6), hx(d, 0x8013BA6, 2))
+
+    d[o:o + 2] = bytes.fromhex('ff20')
+
+    # b40: S12-wrapper pop RESTORED (revert b25 chain at this site).
+    # b25 sent pop to bx ip because the pop read scratch; with S12
+    # preserving sp the frame is intact, while bx ip orphaned 8B per
+    # call and smashed sp into text at 105M. S16 keeps serving B66.
+    o = off(0x8007B66)
+
+    assert d[o:o + 2].hex() == '10bd', hx(d, 0x8007B66, 2)
+
+    # Clock S11 (see b24).
+    o = off(0x800E988)
+
+    assert d[o:o + 2].hex() == '10bd', hx(d, 0x800E988, 2)
+
+    d[o:o + 2] = bytes.fromhex('7047')
+
+    # Poll re-init skip (see b27).
+    o = off(0x8007748)
+
+    assert d[o:o + 2].hex() == '96b1', hx(d, 0x8007748, 2)
+
+    d[o:o + 2] = bytes.fromhex('14e0')
+
+    # Poll-epilogue add NOP (see b28).
+    o = off(0x8007794)
+
+    assert d[o:o + 2].hex() == '07b0', hx(d, 0x8007794, 2)
+
+    d[o:o + 2] = bytes.fromhex('c046')
+
+    # Partition wrap-check (see b33).
+    o = off(0x801844E)
+
+    assert d[o:o + 2].hex() == '04d3', hx(d, 0x801844E, 2)
+
+    d[o:o + 2] = bytes.fromhex('04e0')
+
+    # Hang-site neutralize (see b34).
+    o = off(0x8018458)
+
+    assert d[o:o + 2].hex() == 'fee7', hx(d, 0x8018458, 2)
+
+    d[o:o + 2] = bytes.fromhex('ffe7')
+
+    # Clock-store NOP: str r1,[r3] (2B 1960) -> mov r8,r8 (2B c046).
+    o = off(0x800E986)
+
+    assert d[o:o + 2].hex() == '1960', hx(d, 0x800E986, 2)
+
+    d[o:o + 2] = bytes.fromhex('c046')
+
+    save(d, 'img0_b40.elf')
+
+
 
 
 

tools/sbl_uni.py

@@ -30,6 +30,7 @@ RAMS = [
     (0x00400000, 0x07400000),   # RPM range (heap-ish)
     (0x08000000, 0x00600000),   # DDRLOW
     (0x08600000, 0x00010000),   # IMEM
+    (0x08610000, 0x00FF0000),   # DDRHI (QEMU parity; IMEM-stack spillover)
     (0x80000000, 0x07C00000),   # DDRHIGH_A
     (0x87C00000, 0x00800000),   # SMEM(+SBL BSS)
     (0x07800000, 0x000AF000),   # leak-sink HI (S12 frame march;
@@ -350,25 +351,11 @@ def main():
             except Exception:
                 pass
     mu.hook_add(UC_HOOK_MEM_WRITE, hook_txt, begin=0x802F600, end=0x802F700)
-    _splog2 = []
-    def hook_spck(mu, addr, size, data):
-        try:
-            from unicorn.arm_const import UC_ARM_REG_SP as _S
-            sp = mu.reg_read(_S)
-        except Exception:
-            return
-        if not (0x07800000 <= sp < 0x08610000):
-            if len(_splog2) < 5:
-                _splog2.append((count[0], addr, sp))
-    mu.hook_add(UC_HOOK_CODE, hook_spck)
     import atexit as _ax3
     def _dtx():
         print('txt-writes:', len(_txtlog))
         for n, pc, a, sz, v, sp, lr in _txtlog:
             print('  insn', n, hex(pc), hex(a), sz, hex(v), hex(sp), hex(lr))
-        print('sp-trips:', len(_splog2))
-        for n, pc, sp in _splog2:
-            print('  insn', n, hex(pc), hex(sp))
     _ax3.register(_dtx)
     # null-fetch redirect: branch-to-0 lands in maze (movs r0,#0; bx lr),
     # immune to [0]-writes that clobber a static maze copy