[Bug] Vault accepts a value it later refuses to inject, and the audit trail cannot separate refused from absent
Los mantenedores suelen responder en 1 día
Nadie ha tomado este issue todavía.
Evaluación
- Dificultad
- 4/5
- Tiempo estimado
- 3-5 días
- Aptitud para principiantes
- 52/100
- Tipo de issue
- Error
- Claridad
- Bastante claro
- Estado de actividad
- Activo
- Stack tecnológico
- go
- Área
- authorization, cli, security
Línea de trabajo
Start with the fs-vault create-secret, replace-secret, run-with-secret, read-secret, and list-audit-events command paths, then reproduce both cases from the issue. Check where field values are validated and where delegated reads become audit events. Done means invalid values are refused on write and the audit trail distinguishes denied access to an existing field from a missing field.
Escrito por el modelo de indexación a partir del texto del issue.
Descripción
ti version
0.2.7 (6c950c77a601d0d55394c4b2f64483ae93d522db, 2026-09-27), bundled ti-drive9 companion 60b63d6
Command family
fs-vault
Operating system and architecture
macOS 14.6, darwin/arm64, for both problems. Problem 1 was also seen on Ubuntu, linux/amd64.
Summary
One validation problem and one question about the audit trail.
Covers observations 7 and 4 from a review of the TiDB Cloud Filesystem documentation.
Problem 1: a control byte is accepted on write and refused on use
Command
ti fs-vault create-secret --secret-name ctl --field K="$(printf 'ab\001cd')"
ti fs-vault run-with-secret --secret-path /n/vault/ctl -- /bin/sh -c 'echo injected'
Reproduction steps
- Create or replace a secret whose field value contains a control byte.
- Observe that the command reports success.
- Run
run-with-secretto inject that field.
Expected behavior
The write is refused, at the moment the value is supplied, naming the offending byte.
Actual behavior
create-secret and replace-secret both accept the value and report success at exit 0.
run-with-secret then refuses to inject it at exit 1:
ti [ERROR]: vault with: secret value for key "K" contains forbidden control byte 0x01 at offset 2
A secret with a clean value runs normally through the same command, which confirms the refusal
is about the stored value rather than the invocation. The secret is stored in a state where it cannot
be used, and the caller only discovers this at the point of use, which may be much later and
in a different place.
Validating at write time would remove the failure entirely. As it stands the documentation has
to warn readers about a value the product accepted without complaint, which is a weaker
remedy.
Problem 2: a delegated read cannot be told apart from a missing field
Command
ti fs-vault create-grant --scope db/DB_URL --permission read --agent-id a1 --ttl 30m --token-only
ti fs-vault read-secret --secret-name db --field PASSWORD --vault-token "<delegated-token>"
ti fs-vault read-secret --secret-name db --field NOPE --vault-token "<delegated-token>"
ti fs-vault list-audit-events --output text
Reproduction steps
- Grant a delegated token access to one field of a secret.
- Read a different field of that secret with it.
- Read a field that does not exist at all.
- Compare the messages and the audit entries.
Actual behavior
Both cases return vault get: not found at exit 1, and both record secret.denied in the audit
trail. The granted field reads normally at exit 0.
DB_URL exit=0
PASSWORD exit=1 vault get: not found
NOPE exit=1 vault get: not found
TIME AGENT ACTION SECRET FIELD
2026-09-28T03:42:11Z a1 secret.denied db NOPE
2026-09-28T03:42:11Z a1 secret.denied db PASSWORD
2026-09-28T03:42:11Z a1 secret.read db DB_URL
Note that NOPE appears in the FIELD column as though it were a real field of the secret.
Hiding existence from the caller is a common and defensible vault design, so the message may
well be deliberate. The part worth asking about is the audit trail. An auditor reading
secret.denied cannot tell whether someone was refused access to a field that exists or asked
for a field that never existed. Those are different events, and only one of them is worth
investigating.
This is a question rather than a bug report. Please confirm whether the current behaviour is
intended, and if so I will document it as designed.
Sensitive-data confirmation
No API keys, file system or Vault tokens, database credentials, connection strings, SQL, or
customer data appear above. No real secret values are reproduced. File systems used for these
measurements were deleted afterwards.
- Lenguaje dominante
- Go
- Estrellas
- 3
- Forks
- 4
- Merge medio
- 3 d 19 h
- PR fusionados (30 d)
- 5
Preparar el entorno
Este proyecto no incluye contenedor de desarrollo, Dockerfile ni guía de contribución, así que la configuración corre por tu cuenta: empieza por su README y consulta nuestra guía para la primera contribución para los pasos generales.
Primeros pasos
- Lee el issue completo y luego la guía de contribución del proyecto.
- Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
- Haz un fork del repositorio y trabaja en una rama.
- Abre un pull request que haga referencia al número del issue.
Más de tidbcloud/ti-cli
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 76/100
Los mantenedores suelen responder en 1 día
-
Dificultad 4/5 3-5 días Aptitud para principiantes 48/100
Los mantenedores suelen responder en 1 día
-
Dificultad 5/5 Más de una semana Aptitud para principiantes 25/100
Los mantenedores suelen responder en 1 día
-
Dificultad 4/5 3-5 días Aptitud para principiantes 48/100
Los mantenedores suelen responder en 1 día
-
Dificultad 4/5 3-5 días Aptitud para principiantes 55/100
tidbcloud/ti-cli#64 · 1 comentario ·
Los mantenedores suelen responder en 1 día
Todos los issues de tidbcloud/ti-cli
Issues similares
-
area: metrics area: tracing contribfest enhancement good first issue
Dificultad 2/5 1-3 horas Aptitud para principiantes 83/100
open-telemetry/opentelemetry-ebpf-instrumentation#4188 ·
Los mantenedores suelen responder en 1 día
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 74/100
blinklabs-io/dingo#6074 ·
Los mantenedores suelen responder en 1 día
-
kind/bug remote
Dificultad 2/5 1-3 horas Aptitud para principiantes 68/100
podman-container-tools/podman#30143 ·
Los mantenedores suelen responder en 1 día
-
component/mesheryctl kind/bug language/go
Dificultad 1/5 Menos de una hora Aptitud para principiantes 88/100
meshery/meshery#22993 · 2 comentarios ·
Los mantenedores suelen responder en 1 día
-
enhancement
Dificultad 2/5 1-3 horas Aptitud para principiantes 64/100