Add tar builtins#270
Merged
Merged
Conversation
Follow-up to the tar builtins: an adversarial sweep of the archive code surfaced hang and path-escape vectors. Fix them, and mirror the applicable fixes to the existing zip extractor so the two stay in lockstep. Each case is covered by a test (Tar_test.go, Tar_unix_test.go, ZipHardening_test.go). - Never hang: tarPack rejects fifo/device/socket entries before os.Open, and packing never follows source symlinks, so a symlink loop or a link to /dev/zero cannot block or inflate the archive. - Write-through a pre-existing symlink in the destination is now refused: ensureRealParentWithinBase resolves the deepest existing ancestor via EvalSymlinks and rejects when it lands outside the destination root, while still allowing symlinks that stay inside it. This closes a path-traversal vector present in both the tar and zip extractors. - Decompression-bomb cap: new optional maxBytes key (int, default 0 = unlimited) on the zip/tar extract option dicts bounds the total uncompressed bytes written per extraction via a shared streaming byteBudget. Verified against symlink loops, fifo packing, path traversal, escaping and pre-existing symlinks, hardlinks, truncated/garbage/empty archives, PAX/GNU long names, and a 200MB gzip bomb. Docs and CHANGELOG updated. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reviewed GNU tar's extract.c and its security history for hardening we lack. Its open_output_file / maybe_recoverable logic never writes through a symlink at the target name: it creates with O_EXCL by default and, when overwriting, uses O_NOFOLLOW or unlinks the existing name before recreating. We previously created regular files with O_CREATE|O_WRONLY|O_TRUNC after a stat()-based existence check. Our pre-existing-symlink guard only resolved the parent directory, so with overwrite enabled, extracting an entry whose destination name was already a symlink to an outside file wrote THROUGH it and clobbered the target (verified: dest/victim -> ../secret was overwritten). Route both the tar and zip regular-file writers through createExtractedFile, which opens O_CREATE|O_WRONLY|O_EXCL (never follows a symlink, even a dangling one), and for overwrite removes the existing name itself (os.Remove unlinks the link, not its target) then creates a fresh file atomically. skipExisting and default-error semantics are preserved. Adds regression tests for tar and zip. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Smaller GNU-tar-derived hardenings for the extraction path: - Restore permissions via the open file descriptor (outFile.Chmod) instead of os.Chmod(path), so a metadata restore can never be redirected through a symlink at the destination name. - Reject an archive entry name or symlink target containing an embedded NUL byte before it is used as a path (belt-and-suspenders: Go's archive/tar refuses to encode a NUL, and the OS rejects NUL paths, but a hand-crafted PAX record could carry one). - Add regression tests for escaping symlink targets (absolute and ../ forms) and a direct test of the NUL guard. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GNU tar's decade of symlink-race fixes concluded that lexical/at-check-time containment is inherently TOCTOU-prone, and it now opens every extraction path via openat2(RESOLVE_BENEATH) so the kernel refuses escapes atomically (CVE-2025-45582). Go 1.24+ exposes the same mechanism as os.Root. Route all tar and zip extraction writes (extract-all, single-entry, and subtree) through an os.Root anchored at the destination. Every Mkdir/OpenFile/ Symlink/Chmod/Remove is resolved relative to the root fd, so the kernel blocks any path that escapes via ".." or a symlink component, race-free, while still permitting symlinks that stay within the destination (verified). The prior lexical normalization + containment check is kept as a cheap first layer, and regular files are still created with O_EXCL (no write-through at the final component). Removes the now-superseded ensureRealParentWithinBase EvalSymlinks guard and the non-root createExtractedFile. Behavior is unchanged for valid archives; the full test suite and the adversarial cases (traversal, escaping and pre-existing symlinks, hardlinks, final-component symlink overwrite, in-dest symlink) all pass. 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.
No description provided.