Skip to content

fix(supervisor): reject unsupported live filesystem policy changes #5327

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
nv-rahool wants to merge 1 commit into
base: windows
Choose a base branch
from
Open

fix(supervisor): reject unsupported live filesystem policy changes #5327

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
233 changes: 211 additions & 22 deletions crates/openshell-supervisor/src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3090,6 +3090,34 @@ async fn reload_gateway_policy_runtime(
.await
}

fn validate_live_filesystem_policy(
applied: Option<&openshell_core::proto::SandboxPolicy>,
candidate: Option<&openshell_core::proto::SandboxPolicy>,
) -> Result<()> {
let (Some(applied), Some(candidate)) = (applied, candidate) else {
return Ok(());
};
let same_paths = |left: &[String], right: &[String]| {
left.iter().collect::<std::collections::HashSet<_>>() == right.iter().collect()
};
let unchanged = match (&applied.filesystem, &candidate.filesystem) {
(Some(applied), Some(candidate)) => {
applied.include_workdir == candidate.include_workdir
&& same_paths(&applied.read_only, &candidate.read_only)
&& same_paths(&applied.read_write, &candidate.read_write)
}
(None, None) => true,
_ => false,
};
if !unchanged {
// Reloading the network engine cannot change installed filesystem permissions.
return Err(miette::miette!(
"filesystem policy cannot be changed on a running sandbox; recreate the sandbox to apply filesystem changes"
));
}
Ok(())
}

async fn reload_gateway_configuration_runtime(
engine: &OpaEngine,
policy: Option<&openshell_core::proto::SandboxPolicy>,
Expand Down Expand Up @@ -4681,28 +4709,36 @@ async fn run_policy_poll_loop_with_client<C: PolicyGatewayClient>(

if policy_runtime_changed {
let pid = ctx.entrypoint_pid.load(Ordering::Acquire);
let runtime_result = reload_gateway_configuration_runtime(
&ctx.opa_engine,
let runtime_result = match validate_live_filesystem_policy(
current_endpoint_policy.as_ref(),
result.policy.as_ref(),
pid,
MiddlewareReloadContext {
desired_services: &result.supervisor_middleware_services,
authentication: &MiddlewareAuthentication {
credentials: middleware_credentials.clone(),
enabled: result.extension_authentication_enabled,
},
registry_changed: middleware_registry_changed,
connector: &ctx.middleware_connector,
},
ctx.transparent_tcp,
ctx.vm_identity,
|| {
if let Some(prepared) = prepared_provider.as_ref() {
ctx.provider_credentials.install_prepared(prepared);
}
},
)
.await;
) {
Ok(()) => {
reload_gateway_configuration_runtime(
&ctx.opa_engine,
result.policy.as_ref(),
pid,
MiddlewareReloadContext {
desired_services: &result.supervisor_middleware_services,
authentication: &MiddlewareAuthentication {
credentials: middleware_credentials.clone(),
enabled: result.extension_authentication_enabled,
},
registry_changed: middleware_registry_changed,
connector: &ctx.middleware_connector,
},
ctx.transparent_tcp,
ctx.vm_identity,
|| {
if let Some(prepared) = prepared_provider.as_ref() {
ctx.provider_credentials.install_prepared(prepared);
}
},
)
.await
}
Err(error) => Err(GatewayRuntimeReloadError::PolicyValidation(error)),
};

match runtime_result {
Ok(generation) => {
Expand Down Expand Up @@ -8345,8 +8381,11 @@ network_policies:
1,
openshell_core::proto::PolicySource::Sandbox,
);
// Keep the applied filesystem grants so only the TCP expansion differs.
let mut tcp_policy = proto_policy_fixture();
tcp_policy.network_policies = proto_tcp_policy_fixture().network_policies;
let v2 = settings_poll_result(
Some(proto_tcp_policy_fixture()),
Some(tcp_policy),
2,
openshell_core::proto::PolicySource::Sandbox,
);
Expand Down Expand Up @@ -8386,6 +8425,156 @@ network_policies:
handle.abort();
}

#[tokio::test]
async fn live_filesystem_changes_reject_polled_policy_without_advancing_active_revision() {
use openshell_core::proto::{FilesystemPolicy, PolicySource};

for source in [PolicySource::Sandbox, PolicySource::Global] {
for change in ["read_only", "read_write", "from_none"] {
let mut policy = proto_policy_fixture();
if change == "from_none" {
policy.filesystem = None;
} else {
// Startup may already have added the proxy's filesystem prerequisites.
policy
.filesystem
.as_mut()
.unwrap()
.read_only
.push("/proxy-baseline".into());
}
let initial = settings_poll_result(Some(policy.clone()), 1, source);
let engine = Arc::new(OpaEngine::from_proto(&policy).expect("build OPA engine"));
let mut ctx = policy_poll_test_context(
engine.clone(),
LoadedPolicyOrigin::Gateway {
revision: Some(LoadedPolicyRevision::from_snapshot(&initial)),
has_last_valid_policy: true,
},
default_middleware_connector(),
);
ctx.endpoint_policy = Some(policy.clone());
let (client, polls, mut reports) = scripted_policy_gateway();
let observed_polls = client.polled_sandboxes.clone();
let wait_for_polls = |count: usize| {
let observed_polls = observed_polls.clone();
async move {
timeout(Duration::from_secs(1), async {
while observed_polls.lock().await.len() < count {
tokio::task::yield_now().await;
}
})
.await
.expect("policy poll loop stalled");
}
};
polls.send(initial).unwrap();
let handle = tokio::spawn(run_policy_poll_loop_with_client(ctx, client));
if source == PolicySource::Sandbox {
expect_policy_report(&mut reports, 1).await;
}
wait_for_polls(2).await;
let original_generation = engine.current_generation();

let filesystem = policy
.filesystem
.get_or_insert_with(FilesystemPolicy::default);
if change == "read_write" {
filesystem.read_write.push("/extra-readwrite".into());
} else {
filesystem.read_only.push("/extra-readonly".into());
}
let mut candidate = settings_poll_result(Some(policy), 2, source);
let fail_closed = source == PolicySource::Global && change == "read_only";
candidate.policy_validation_failure_mode = if fail_closed {
PolicyValidationFailureMode::FailClosed
} else {
PolicyValidationFailureMode::RetainLastValid
};
polls.send(candidate).unwrap();
wait_for_polls(3).await;

if source == PolicySource::Sandbox {
let (version, loaded, error) = timeout(Duration::from_secs(1), reports.recv())
.await
.expect("filesystem rejection report timed out")
.expect("policy reporter stopped");
assert_eq!(version, 2, "{source:?}: {change}");
assert!(!loaded, "{source:?}: {change}");
assert!(
error.contains("filesystem"),
"{source:?}: {change}: {error}"
);
} else {
expect_no_policy_report(&mut reports).await;
}
assert_eq!(
engine.fail_closed_reason().is_some(),
fail_closed,
"{source:?}: {change}"
);
assert_eq!(
engine.current_generation(),
original_generation + u64::from(fail_closed),
"{source:?}: {change}"
);
handle.abort();
}
}
}

#[tokio::test]
async fn live_filesystem_change_poll_reports_failure_then_accepts_network_only_repair() {
use openshell_core::proto::PolicySource;

let policy = proto_policy_fixture();
let initial = settings_poll_result(Some(policy.clone()), 1, PolicySource::Sandbox);
let mut rejected = policy.clone();
rejected
.filesystem
.as_mut()
.unwrap()
.read_only
.push("/extra-readonly".into());
let mut rejected = settings_poll_result(Some(rejected), 2, PolicySource::Sandbox);
rejected.policy_validation_failure_mode = PolicyValidationFailureMode::RetainLastValid;
let mut repaired = policy.clone();
repaired.network_policies = proto_tcp_policy_fixture().network_policies;
repaired.filesystem.as_mut().unwrap().read_only.reverse();
let repaired = settings_poll_result(Some(repaired), 3, PolicySource::Sandbox);
let engine = Arc::new(OpaEngine::from_proto(&policy).expect("build OPA engine"));
let mut ctx = policy_poll_test_context(
engine.clone(),
LoadedPolicyOrigin::Gateway {
revision: Some(LoadedPolicyRevision::from_snapshot(&initial)),
has_last_valid_policy: true,
},
default_middleware_connector(),
);
ctx.endpoint_policy = Some(policy);
ctx.transparent_tcp = TransparentTcpReloadState {
capable: true,
substrate_ready: true,
};
let (client, polls, mut reports) = scripted_policy_gateway();
polls.send(initial).unwrap();
let handle = tokio::spawn(run_policy_poll_loop_with_client(ctx, client));
expect_policy_report(&mut reports, 1).await;
polls.send(rejected).unwrap();
let (version, loaded, error) = timeout(Duration::from_secs(1), reports.recv())
.await
.expect("filesystem rejection report timed out")
.expect("policy reporter stopped");
assert_eq!(version, 2);
assert!(!loaded);
assert!(error.contains("filesystem"));
assert_eq!(engine.current_generation(), 0);
polls.send(repaired).unwrap();
expect_policy_report(&mut reports, 3).await;
assert_eq!(engine.current_generation(), 1);
handle.abort();
}

#[tokio::test]
async fn same_hash_ack_waits_for_failed_middleware_reconciliation_and_retries_once() {
let mut v1 = settings_poll_result(
Expand Down
2 changes: 1 addition & 1 deletion docs/how-it-works/policies/manage-policies.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -191,7 +191,7 @@ running sandbox. The startup sections do not:
| Network rules | The sandbox loads the new rules. Connections opened under the previous rules close. | Apply the change, then retry the request. |
| Middleware in `network_middlewares` | You can add, remove, or reconfigure built-in middleware and external services already registered with the gateway. | Apply the change with `openshell policy set`. |
| External middleware registration | Policy changes cannot register a service or change its gateway connection settings. | Update the gateway configuration and restart the gateway. |
| Added filesystem paths | The saved policy can change, but the running workload keeps its existing filesystem permissions. | Recreate the sandbox. |
| Added filesystem paths | The sandbox rejects the entire revision, including network changes. With the default `fail_closed` failure mode, it blocks network traffic until a valid revision loads. | Recreate the sandbox. |
| Removed filesystem paths, or changed `include_workdir`, Landlock, or process settings | OpenShell rejects the change after the workload starts. | Recreate the sandbox. |

When new network rules take effect, OpenShell closes connections that were
Expand Down
Loading