Unlocks the x86 family end-to-end and makes the write-loop detector precise
enough for statically-linked binaries.
x86 SLEIGH:
- Vendor Ghidra's x86 language specs (specs/x86, x86:LE:64:default + :32:default);
add x86-64/x86 .sla build targets. archSpecFor maps x86/64 + x86/32.
- ghidra_lifter: set the x86-64 decode context (longMode/addrsize/opsize) in the
raw path, and bound decompiler flow at the function end (0xC3 tail fill) so
variable-length x86 does not run away into zero padding.
Precise SSA def-use (benefits all arches):
- ir::pcode::PcodeOp gains in_def (defining-op index per input); the SSA lifter
records real def-use. summarizePointerWrites now resolves the store-address
base via def-use instead of a last-write heuristic — essential on x86 where
arg registers are reused as scratch (RAX) and sub-registers are written (SIL).
- paramIndexOf is arch-aware (AArch64 x0–x7; SysV x86-64 RDI/RSI/RDX/RCX/R8/R9).
Precision (the key fix at scale):
- A copy bounded by a size *parameter* (memcpy-like `for i<n`) is suppressed:
if the loop index is `<`-compared against a pure parameter input, the function
is size-bounded and not flagged. baby_bof's decode_base64 bounds on
strlen(input) (a computed value, not a param) so it still fires. This cut a
statically-linked x86-64 binary from 101 findings (≈100 glibc/libstdc++ FPs)
to 1 — the real decode_base64 → 264-byte overflow.
Perf + reach:
- Cache the pre-pass SSA lift and reuse it in the main pass (re-lifting only the
few callers of a write-loop callee that need the injected prototype).
- Give whole-binary r2 analysis a generous timeout (FOLD_R2_ANALYSIS_TIMEOUT_MS,
default 120s) so large static binaries actually analyze (the 3s per-command
watchdog left them with an empty function list).
Validated: decode_base64 overflow fires on x86-64 (index_amd64.cgi) and AArch64
(index_nocanary.cgi); AArch64 corpus BOF counts unchanged (3/1/1); x86-64 static
FP 101→1. Unit: test_pcode_summary +x86-64 case (9 assertions). meson test 77/77.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Foundation for replacing fold's dead/abstract binary encoder with a real
SSA P-Code frontend. Verified root-cause gap: serializeBinaryIR() emits only
data-discovery nodes (sections/strings/imports/exports); the control-flow
BinaryExprKind values (BasicBlock/CallSite/...) are never constructed, so the
Z3-BB encoder's encodable filter is always empty and encodeBinaryBBs never
runs — hence 0 binary Z3 findings. The P-Code pipeline supersedes that path.
Phase 0 — docs/internal/ghidra-pcode-integration-research.md: P-Code opcode
core, the decompiler C++ embedding API (LoadImage/SleighArchitecture/Funcdata),
the CORE+DECCORE+SLEIGH source sets, no-bison finding, and pypcode as build
reference.
Phase 1 — vendored Ghidra decompiler at pinned commit (Ghidra_12.1_build,
7e89d94e34...) under third_party/ghidra-decompiler/ (cpp + AArch64 specs +
Apache-2.0 LICENSE + VENDOR.md + .gitignore for build artifacts). Validated by
hand on this clang-21/macOS toolchain:
- full SSA engine (CORE+DECCORE+SLEIGH, 80 files incl. heritage/funcdata/
action) compiles clean, -std=c++11 -O2, NO BFD, NO bison (generated
parsers are committed upstream; ruleparse isn't in the lib set);
- sleigh compiler builds; AArch64 .slaspec -> .sla compiles;
- tools/pcode_dump.cpp (custom LoadImage, no BFD) lifts the real arm64
sym._div_by_zero_direct to correct P-Code: `mov w9,#0 -> COPY(const 0)`,
`sdiv -> INT_SDIV(w8,w9)` — divisor zero-ness explicit in the data flow,
so the Phase-3 encoder proves div-by-zero trivially, multi-arch.
Not yet meson-wired (next step) and no fold-side IR/encoder yet (Phase 2-3);
this commit is the validated vendor + smoke tool.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
