Harden backup restore tar extraction #574

Merged

ghreprimand merged 1 commit from ghreprimand/harden-backup-restore-tar into main

2026-06-01 20:08:46 +02:00

ghreprimand commented

2026-06-01 16:54:34 +02:00

(Migrated from github.com)

Summary

validate backup archive members before restore and verify
reject symlink, hardlink, and special-file entries instead of passing them to tar.extractall()
extract only regular files and directories under data/ manually
skip symlinked files when creating snapshots so generated backups match the hardened restore contract

Why

The restore path already rejected absolute paths and ../ traversal, but tar archives can also redirect writes through link entries. For example, an archive can create a link inside data/ and then include a later regular file beneath that link. Using extractall() leaves that class of archive behavior to tarfile.

This keeps restores constrained to the data/ tree by accepting only regular files/directories and copying file contents directly after validation. It also makes odysseus-backup verify reject archives that restore would refuse.

Tests

venv/bin/python -m pytest tests/test_backup_cli_security.py -q
venv/bin/python -m pytest tests/test_security_regressions.py -q
venv/bin/python -m py_compile scripts/odysseus-backup
git diff --check

## Summary - validate backup archive members before restore and verify - reject symlink, hardlink, and special-file entries instead of passing them to tar.extractall() - extract only regular files and directories under data/ manually - skip symlinked files when creating snapshots so generated backups match the hardened restore contract ## Why The restore path already rejected absolute paths and ../ traversal, but tar archives can also redirect writes through link entries. For example, an archive can create a link inside data/ and then include a later regular file beneath that link. Using extractall() leaves that class of archive behavior to tarfile. This keeps restores constrained to the data/ tree by accepting only regular files/directories and copying file contents directly after validation. It also makes odysseus-backup verify reject archives that restore would refuse. ## Tests - venv/bin/python -m pytest tests/test_backup_cli_security.py -q - venv/bin/python -m pytest tests/test_security_regressions.py -q - venv/bin/python -m py_compile scripts/odysseus-backup - git diff --check

sleepy added the

bug

label

2026-06-01 20:00:42 +02:00

sleepy merged commit d8451ea849 into main

2026-06-01 20:08:46 +02:00

sleepy referenced this pull request from a commit

2026-06-01 20:08:46 +02:00

Harden backup restore tar extraction (#574)

sleepy commented

2026-06-01 20:08:47 +02:00

Owner

Merged via squash. Restore path now validates members before extraction, rejects symlinks/hardlinks/special files, and uses safe member-by-member extraction instead of extractall(). All 4 tests pass.