GHSA-8W8G-WQ8H-FQ33
Vulnerability from github – Published: 2026-10-02 18:53 – Updated: 2026-10-02 18:53Summary
Dulwich's porcelain.checkout(paths=[...]) code path writes files using raw os.open(file_path, O_WRONLY|O_CREAT|O_TRUNC, mode) followed by f.write(obj.data). This code path does NOT call build_file_from_blob() at all, completely bypassing any symlink protections (including the unreleased d09f8af fix). os.open without O_NOFOLLOW follows symlinks at both the target file and intermediate directories, allowing arbitrary file writes.
Root Cause
At dulwich/porcelain/__init__.py:5661-5675, the checkout(paths=[...]) implementation:
file_path = _checked_worktree_path(r, path)
os.makedirs(os.path.dirname(file_path), exist_ok=True)
flags = os.O_WRONLY | os.O_CREAT | os.O_TRUNC
with os.fdopen(os.open(file_path, flags, mode), "wb") as f:
f.write(obj.data)
_checked_worktree_path() (line 601-631) only performs name validation — checking that the path doesn't start with / or \\ and that components pass INVALID_DOTNAMES checks. It performs zero filesystem symlink detection.
Impact
An attacker can craft a malicious repository that, when a victim clones it and runs checkout(paths=[...]), writes attacker-controlled content (with attacker-controlled permissions) to any filesystem location accessible to the user. Writing to .git/hooks/post-checkout achieves RCE on the next git checkout.
Attack Scenario
- Attacker creates a repository where HEAD has
triggeras a symlink (mode 120000, content../../.git/hooks/post-checkout), and tagv1.0hastriggeras an executable file (mode 100755, content#!/bin/sh\nmalicious_payload) - Victim clones the repository — worktree has
trigger→../../.git/hooks/post-checkout(a symlink) - Victim runs
porcelain.checkout(repo, target="v1.0", paths=["trigger"])to restore a specific file from a tag _checked_worktree_path(r, "trigger")passes — name validation only, no symlink checkos.open("trigger", O_WRONLY|O_CREAT|O_TRUNC, 0o755)follows the symlink → opens.git/hooks/post-checkoutfor writingf.write(obj.data)writes the malicious payload to the hook- Next checkout operation triggers the hook → RCE
Suggested Fix
Replace the raw os.open path with a call to build_file_from_blob (once that function is hardened against intermediate symlinks), or add explicit symlink detection: resolve the path with os.path.realpath() and verify it stays within the worktree root before opening.
Reported by zx (Jace)
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 1.2.7"
},
"package": {
"ecosystem": "PyPI",
"name": "dulwich"
},
"ranges": [
{
"events": [
{
"introduced": "0.24.0"
},
{
"fixed": "1.2.8"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [],
"database_specific": {
"cwe_ids": [
"CWE-59",
"CWE-61"
],
"github_reviewed": true,
"github_reviewed_at": "2026-10-02T18:53:35Z",
"nvd_published_at": null,
"severity": "HIGH"
},
"details": "## Summary\n\nDulwich\u0027s `porcelain.checkout(paths=[...])` code path writes files using raw `os.open(file_path, O_WRONLY|O_CREAT|O_TRUNC, mode)` followed by `f.write(obj.data)`. This code path does NOT call `build_file_from_blob()` at all, completely bypassing any symlink protections (including the unreleased d09f8af fix). `os.open` without `O_NOFOLLOW` follows symlinks at both the target file and intermediate directories, allowing arbitrary file writes.\n\n## Root Cause\n\nAt `dulwich/porcelain/__init__.py:5661-5675`, the `checkout(paths=[...])` implementation:\n\n```python\nfile_path = _checked_worktree_path(r, path)\nos.makedirs(os.path.dirname(file_path), exist_ok=True)\nflags = os.O_WRONLY | os.O_CREAT | os.O_TRUNC\nwith os.fdopen(os.open(file_path, flags, mode), \"wb\") as f:\n f.write(obj.data)\n```\n\n`_checked_worktree_path()` (line 601-631) only performs name validation \u2014 checking that the path doesn\u0027t start with `/` or `\\\\` and that components pass `INVALID_DOTNAMES` checks. It performs zero filesystem symlink detection.\n\n## Impact\n\nAn attacker can craft a malicious repository that, when a victim clones it and runs `checkout(paths=[...])`, writes attacker-controlled content (with attacker-controlled permissions) to any filesystem location accessible to the user. Writing to `.git/hooks/post-checkout` achieves RCE on the next git checkout.\n\n## Attack Scenario\n\n1. Attacker creates a repository where HEAD has `trigger` as a symlink (mode 120000, content `../../.git/hooks/post-checkout`), and tag `v1.0` has `trigger` as an executable file (mode 100755, content `#!/bin/sh\\nmalicious_payload`)\n2. Victim clones the repository \u2014 worktree has `trigger` \u2192 `../../.git/hooks/post-checkout` (a symlink)\n3. Victim runs `porcelain.checkout(repo, target=\"v1.0\", paths=[\"trigger\"])` to restore a specific file from a tag\n4. `_checked_worktree_path(r, \"trigger\")` passes \u2014 name validation only, no symlink check\n5. `os.open(\"trigger\", O_WRONLY|O_CREAT|O_TRUNC, 0o755)` follows the symlink \u2192 opens `.git/hooks/post-checkout` for writing\n6. `f.write(obj.data)` writes the malicious payload to the hook\n7. Next checkout operation triggers the hook \u2192 RCE\n\n## Suggested Fix\n\nReplace the raw `os.open` path with a call to `build_file_from_blob` (once that function is hardened against intermediate symlinks), or add explicit symlink detection: resolve the path with `os.path.realpath()` and verify it stays within the worktree root before opening.\n\nReported by **zx (Jace)**",
"id": "GHSA-8w8g-wq8h-fq33",
"modified": "2026-10-02T18:53:35Z",
"published": "2026-10-02T18:53:35Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/jelmer/dulwich/security/advisories/GHSA-8w8g-wq8h-fq33"
},
{
"type": "WEB",
"url": "https://github.com/jelmer/dulwich/commit/9389fcb5cb9113adfc7f207d8be86a56904db3e8"
},
{
"type": "PACKAGE",
"url": "https://github.com/jelmer/dulwich"
},
{
"type": "WEB",
"url": "https://github.com/jelmer/dulwich/releases/tag/dulwich-1.2.8"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H",
"type": "CVSS_V3"
}
],
"summary": "Dulwich: Symlink write-through in checkout(paths=[]) via raw os.open bypasses all symlink protections"
}
Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
Vulnerability timeline
Sightings and recorded dates, grouped by day · UTC
Loading timeline…
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.