fix(airflow-dag-parser): close FP5 parser stack-overflow DoS (v1.0.1)#2
Merged
Merged
Conversation
Port FerroAir's complete recursion guard into the crates.io-published `ferro-airflow-dag-parser` and harden it via a Codex DD + Claude-critic convergence loop. Parsing attacker-controlled Python could overflow the vendored `littrs-ruff-python-parser` 0.6.2 recursive descent and abort the process with a SIGSEGV (`catch_unwind` cannot intercept a guard-page fault). Three-layer guard in panic_safe.rs: - Layer 1: iterative bracket pre-scan (cap 256). - Layer 1b: single real-tokenizer pass bounding `brackets + op_run + line_right_rec + indent` (cap 1024); rejects PEP-750 t-strings. Counted run includes the non-bracket recursion vectors: not/await/~/-/+ prefix, a**b**c, a if b else / lambda, mixed prefix chains, yield/yield-from, and async error-recovery chains. - Layer 2: parse AND AST walk run on a dedicated 128 MiB stack; thread-join folds any unwinding panic into ParseError::Internal. Recursive AST walkers (collect_shift_edges, stringify_expr, resolve_to_task_id, inner_name, marker decorator helpers) additionally truncate past depth 1024 so a deep left-leaning >>/attr/call chain that survives the parser cannot overflow the walk either. Adversarial convergence (Codex DD 10 rounds + 10 Claude-critic empirical passes, ~240 shapes; full 110-variant TokenKind classification; fuzz 1.15M runs / 0 crash): fixed stringify-walker, yield/from, async, NonLogicalNewline/Comment trivia-reset, and marker decorator-chain recursion; R9+R10 NO FINDINGS. Honest residual (dd-pack/11-known-limitations.md): a single left-leaning trailer chain in a multi-MB file can still overflow on recursive AST construction/drop (Rust Drop, shared with ferro-air, 128 MiB-bounded); not closable pre-parse without false-positives. Realistic FP5 shapes (<4 KiB) are fully closed. No public-API change; additive, semver-compatible. clippy -D warnings clean, 170 tests, fmt/deny/SPDX/MSRV 1.88/forbid(unsafe_code) green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes FP5 — the parser stack-overflow DoS in the crates.io-published
ferro-airflow-dag-parser. Published as v1.0.1.What
Parsing attacker-controlled Python could overflow the vendored
littrs-ruff-python-parser0.6.2 recursive descent and abort the host process with aSIGSEGV(catch_unwindcannot intercept a guard-page fault). Ports FerroAir's complete three-layer recursion guard and hardens it via an adversarial convergence loop.Guard (panic_safe.rs)
brackets + op_run + line_right_rec + indent(cap 1024); rejects PEP-750 t-strings. Counted run covers the non-bracket vectors:not/await/~/-/+prefix,a**b**c,a if b else/lambda, mixed prefix,yield/yield from, andasyncerror-recovery chains.ParseError::Internal.stringify_expr,collect_shift_edges,resolve_to_task_id, decorator helpers) truncate past depth 1024.Convergence (Codex DD + Claude-critic)
10 Codex rounds + 10 empirical passes (~240 shapes) + full 110-variant
TokenKindclassification + fuzz ~2.66M runs / 0 crash. 5 fixes: stringify walker,yield/from,async,NonLogicalNewline/Commenttrivia-reset (a latent bug shared with ferro-air's lex_reject), marker decorator-chain. R9+R10 NO FINDINGS.Honest residual
A single left-leaning trailer chain in a multi-MB file can still overflow on recursive AST construction/drop (Rust
Drop, shared with ferro-air, 128 MiB-bounded; not closable pre-parse without false-positives). Realistic FP5 shapes (<4 KiB) are fully closed. Seedd-pack/11-known-limitations.md.Quality
No public-API change; additive, semver-compatible.
clippy -D warningsclean, 170 tests, fmt/deny/SPDX/MSRV 1.88/forbid(unsafe_code)green.🤖 Generated with Claude Code