Policy selection is nondeterministic when several policies allow one image; add container_names routing and deny on ambiguity
Los mantenedores suelen responder en 1 día
Ya se ha fusionado un pull request relacionado.
- #90 de @abienkowski — fusionado
Evaluación
- Dificultad
- 4/5
- Tiempo estimado
- 3-5 días
- Aptitud para principiantes
- 18/100
- Tipo de issue
- Error
- Claridad
- Bastante claro
- Estado de actividad
- Activo
- Stack tecnológico
- docker, go, rust, typescript
Línea de trabajo
Start from GetByImage, which returns the first matching policy while iterating a map, and from the POST /containers/create handler that uses its result. Adding container_names routing and load-time ambiguity checks touches all three implementations and the Quint model. Pull request #90 is already open against this issue, so check it before starting. Done means the same image and name always select one policy, and ambiguous matches return 403.
Escrito por el modelo de indexación a partir del texto del issue.
Descripción
Describe the bug
When more than one loaded policy allows the same image, the policy applied to a create is effectively random. GetByImage returns the first policy whose prefixes match while iterating a map: Go map and Rust HashMap iteration order is randomized, so the choice can change from one request to the next; TypeScript's Map keeps load order, so the first-loaded policy always wins. The selected policy decides the forced container_config.user, the volume whitelist and the flag rules, so a container can be created with another service's settings.
This is a normal layout, not an edge case: services built from one image need different policies. Lodestar's beacon node and validator client both run chainsafe/lodestar with different users and data directories, so in a one-trust-domain deployment (README "Trust model") validator creates intermittently get the beacon policy — forced to the beacon user and refused their own data directory.
Affected implementation(s)
- Go
- Rust
- TypeScript
- All
To Reproduce
- Two policies in one
--config-dir:# a.yaml service_name: svc-a allowed_image_prefixes: [alpine] container_config: { user: "1111:1111" } # b.yaml service_name: svc-b allowed_image_prefixes: [alpine] container_config: { user: "2222:2222" } - Through the proxy,
docker create alpine:3.20 truesix times, restart the proxy, repeat (three runs). - Forced users observed (Go v0.3.11):
run 1: 1111 1111 1111 1111 1111 1111 run 2: 1111 2222 1111 1111 1111 1111 run 3: 1111 1111 1111 1111 1111 1111
Expected behavior
Policy selection is deterministic, and services that share an image can be told apart.
Proposed:
- Optional
container_namesin a policy (e.g.container_names: [beacon]). OnPOST /containers/create, select the policy whose image prefixes match and whosecontainer_namescontains the request's?name=. A policy withoutcontainer_namesmatches any name, as today. - Exactly one match, otherwise deny. Zero matches → 403 as today; more than one → 403 with an ambiguity error naming the matching policies. Validate at load time too: two policies that could both match the same image and name should fail startup.
The name is chosen by the caller, so this is routing, not access control — consistent with the trust model, where every caller of a socket already holds the union of its policies.
Environment
- OS: Ubuntu 22.04.5
- Docker version: Engine 29.9.0 (API 1.56)
- Implementation version/commit: v0.3.11 (
3130b0a), Go binary
Additional context
Needed by ChainSafe/lodestar-infrastructure's deployment of this proxy (beacon and validator from one image). The Quint model would need the container name as an input to state "the applied policy is the unique match for (image, name)".
- Lenguaje dominante
- Rust
- Estrellas
- 2
- Forks
- 0
- Merge medio
- 12 h 11 min
- PR fusionados (30 d)
- 30
Preparar el entorno
- Sin Dockerfile ni archivo de Docker Compose
- Tiene una plantilla de pull request
- Leer la guía de contribución
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 ChainSafe/docker-socket-policy
-
Type: Feature
Dificultad 5/5 Más de una semana Aptitud para principiantes 25/100
ChainSafe/docker-socket-policy#65 ·
Los mantenedores suelen responder en 1 día
-
Type: Enhancement
Dificultad 5/5 Más de una semana Aptitud para principiantes 38/100
ChainSafe/docker-socket-policy#46 · 1 comentario ·
Los mantenedores suelen responder en 1 día
Todos los issues de ChainSafe/docker-socket-policy
Issues similares
-
Four problem loader browser tests fail on WindowsPosiblemente ocupada @Patriciaath999 la tomó hoy. Abierto
Dificultad 2/5 1-3 horas Aptitud para principiantes 88/100
Los mantenedores suelen responder en 1 día
-
Claude Code guards in .claude/settings.json never block: they read unset variables and exit 1Abierto
Dificultad 2/5 1-3 horas Aptitud para principiantes 62/100
starkware-libs/sequencer#15229 ·
Los mantenedores suelen responder en 1 día
-
`fnox set` ignores piped stdin when `--description` is given, saving the secret without a valueAbierto
Dificultad 2/5 1-3 horas Aptitud para principiantes 68/100
Los mantenedores suelen responder en 1 día
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 70/100
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 66/100
ActivityWatch/aw-notify-rs#47 · 1 comentario ·