add one shared templated agent policy and name each agent's bao entity after it

swarm-controller/src/agent_entity.rs

@@ -0,0 +1,299 @@
+//! The identity entity an agent's login lands on, named after the agent.
+//!
+//! A cert login creates an entity with a generated name and an alias named for
+//! the certificate's common name, `hive-agent-<agent>`. The shared policy
+//! ([`policy::render_shared_agent`]) reads `{{identity.entity.name}}`, so the
+//! entity has to be called `<agent>`. This module makes it so, either by
+//! creating the entity and its alias before the first login or by renaming the
+//! entity a login already created. The alias is unique per auth mount and
+//! name, so there is never a second entity to pick between.
+//!
+//! An entity's name is its access under the shared policy, so only an entity
+//! the store named itself is ever renamed ([`is_generated_name`]) and no alias
+//! is ever moved. Any other arrangement is a [`Plan::Conflict`]: nothing is
+//! written and an operator decides.
+//!
+//! Everything else here is best effort. An agent's own policy still grants it
+//! its paths, so a store that refuses these calls costs the shared policy and
+//! nothing else: the failure is logged at `warn` and the caller goes on. The
+//! grants are the `identity/` stanzas of `swarm-bao.nix`'s
+//! `controllerPolicyText`, and a controller can start before they deploy.
+
+use anyhow::{Context, Result, bail};
+use swarm_secret_client::{
+    SecretStore,
+    client::{DEFAULT_CERT_MOUNT, EntityRef},
+    policy,
+};
+
+/// What the store holds for one agent's entity.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct Seen {
+    /// The entity that owns the agent's alias on the cert mount.
+    pub by_alias: Option<EntityRef>,
+    /// The id of the entity named after the agent.
+    pub by_name: Option<String>,
+}
+
+/// What to do about it.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum Plan {
+    /// The alias's entity is already named after the agent.
+    Keep,
+    /// A login created the entity under a generated name.
+    Rename { id: String },
+    /// An entity named after the agent exists, and the alias is on no entity.
+    AddAlias { entity_id: String },
+    /// Neither the entity nor the alias exists.
+    CreateAndAlias,
+    /// The alias sits on an entity that is not the agent's: one with a name of
+    /// someone's choosing, or one while another entity already has the
+    /// agent's name. Nothing is written.
+    Conflict { alias: String, entity_name: String },
+}
+
+/// Whether `name` is one the store gave an entity at login: `entity_`, eight
+/// lowercase hex digits, and the root namespace's `.root` suffix, which stores
+/// that predate namespaces leave off.
+///
+/// Exact on purpose. A name that is merely close, such as `entity_` followed
+/// by a word, is somebody's choice of name.
+#[must_use]
+pub fn is_generated_name(name: &str) -> bool {
+    let Some(rest) = name.strip_prefix("entity_") else {
+        return false;
+    };
+    let digits = rest.strip_suffix(".root").unwrap_or(rest);
+    digits.len() == 8
+        && digits
+            .bytes()
+            .all(|b| b.is_ascii_digit() || (b'a'..=b'f').contains(&b))
+}
+
+/// The plan that leaves the alias on an entity named `agent`.
+#[must_use]
+pub fn plan(agent: &str, seen: &Seen) -> Plan {
+    match (&seen.by_alias, &seen.by_name) {
+        (Some(entity), _) if entity.name == agent => Plan::Keep,
+        (Some(entity), None) if is_generated_name(&entity.name) => Plan::Rename {
+            id: entity.id.clone(),
+        },
+        (Some(entity), _) => Plan::Conflict {
+            alias: format!("{}{agent}", policy::AGENT_PREFIX),
+            entity_name: entity.name.clone(),
+        },
+        (None, Some(named)) => Plan::AddAlias {
+            entity_id: named.clone(),
+        },
+        (None, None) => Plan::CreateAndAlias,
+    }
+}
+
+/// Make sure every agent in `agents` has its entity, logging what could not be
+/// done. Never fails: see the module header.
+pub async fn ensure_agents(store: &SecretStore, agents: &[String]) {
+    let accessor = match store.auth_mount_accessor(DEFAULT_CERT_MOUNT).await {
+        Ok(accessor) => accessor,
+        Err(e) => {
+            tracing::warn!(
+                error = %e,
+                "agent entities are not managed: the cert auth mount's accessor could not be read; agents keep their own policy"
+            );
+            return;
+        }
+    };
+    let mut changed = 0_usize;
+    for agent in agents {
+        match ensure_one(store, &accessor, agent).await {
+            Ok(Plan::Keep | Plan::Conflict { .. }) => {}
+            Ok(_) => changed += 1,
+            Err(e) => {
+                tracing::warn!(agent, error = %format!("{e:#}"), "this agent's identity entity could not be set up");
+            }
+        }
+    }
+    tracing::info!(
+        changed,
+        agents = agents.len(),
+        "agent identity entities checked"
+    );
+}
+
+/// [`ensure_agents`] for one agent.
+pub async fn ensure_agent(store: &SecretStore, agent: &str) {
+    ensure_agents(store, &[agent.to_owned()]).await;
+}
+
+/// Observe, plan and apply for one agent.
+async fn ensure_one(store: &SecretStore, accessor: &str, agent: &str) -> Result<Plan> {
+    let alias = policy::agent_object_name(agent)?;
+    let seen = Seen {
+        by_alias: store
+            .entity_by_alias(&alias, accessor)
+            .await
+            .with_context(|| format!("looking up the entity owning alias {alias}"))?,
+        by_name: store
+            .entity_id_by_name(agent)
+            .await
+            .with_context(|| format!("looking up the entity named {agent}"))?,
+    };
+    let plan = plan(agent, &seen);
+    match &plan {
+        Plan::Keep => {}
+        Plan::Rename { id } => {
+            store
+                .rename_entity(id, agent)
+                .await
+                .with_context(|| format!("renaming entity {id} to {agent}"))?;
+            tracing::info!(agent, entity = %id, "agent identity entity renamed");
+        }
+        Plan::AddAlias { entity_id } => {
+            store
+                .create_entity_alias(&alias, accessor, entity_id)
+                .await
+                .with_context(|| format!("adding alias {alias} to entity {entity_id}"))?;
+            tracing::info!(agent, entity = %entity_id, "agent identity alias added");
+        }
+        Plan::CreateAndAlias => {
+            store
+                .create_entity(agent)
+                .await
+                .with_context(|| format!("creating the entity {agent}"))?;
+            let Some(entity_id) = store.entity_id_by_name(agent).await? else {
+                bail!("the entity {agent} was created and is not there to read back");
+            };
+            store
+                .create_entity_alias(&alias, accessor, &entity_id)
+                .await
+                .with_context(|| format!("adding alias {alias} to entity {entity_id}"))?;
+            tracing::info!(agent, entity = %entity_id, "agent identity entity created");
+        }
+        Plan::Conflict { alias, entity_name } => {
+            tracing::error!(
+                agent,
+                alias,
+                entity_name,
+                "the agent's alias is on an entity that is not the agent's; nothing was changed and the agent keeps its own policy"
+            );
+        }
+    }
+    Ok(plan)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    fn entity(id: &str, name: &str) -> EntityRef {
+        EntityRef {
+            id: id.to_owned(),
+            name: name.to_owned(),
+        }
+    }
+
+    fn conflict(entity_name: &str) -> Plan {
+        Plan::Conflict {
+            alias: "hive-agent-atlas".to_owned(),
+            entity_name: entity_name.to_owned(),
+        }
+    }
+
+    #[test]
+    fn an_entity_already_named_after_the_agent_is_kept() {
+        let seen = Seen {
+            by_alias: Some(entity("e1", "atlas")),
+            by_name: Some("e1".to_owned()),
+        };
+        assert_eq!(plan("atlas", &seen), Plan::Keep);
+    }
+
+    #[test]
+    fn nothing_at_all_is_created_with_its_alias() {
+        let seen = Seen {
+            by_alias: None,
+            by_name: None,
+        };
+        assert_eq!(plan("atlas", &seen), Plan::CreateAndAlias);
+    }
+
+    #[test]
+    fn an_entity_a_login_named_is_renamed() {
+        for generated in ["entity_1a2b3c4d.root", "entity_1a2b3c4d"] {
+            let seen = Seen {
+                by_alias: Some(entity("e1", generated)),
+                by_name: None,
+            };
+            assert_eq!(
+                plan("atlas", &seen),
+                Plan::Rename {
+                    id: "e1".to_owned()
+                },
+                "{generated}"
+            );
+        }
+    }
+
+    #[test]
+    fn an_entity_somebody_named_is_never_renamed() {
+        let seen = Seen {
+            by_alias: Some(entity("e1", "bob")),
+            by_name: None,
+        };
+        assert_eq!(plan("atlas", &seen), conflict("bob"));
+    }
+
+    #[test]
+    fn a_name_only_close_to_the_generated_shape_is_somebodys_choice() {
+        for name in [
+            "entity_",
+            "entity_backup",
+            "entity_1a2b3c4",
+            "entity_1a2b3c4d5",
+            "entity_1A2B3C4D",
+            "entity_1a2b3c4g",
+            "entity_1a2b3c4d.root2",
+            "entity_1a2b3c4d.other",
+            "entity_1a2b3c4d.root.root",
+            "xentity_1a2b3c4d",
+            "entity-1a2b3c4d",
+        ] {
+            let seen = Seen {
+                by_alias: Some(entity("e1", name)),
+                by_name: None,
+            };
+            assert_eq!(plan("atlas", &seen), conflict(name), "{name}");
+        }
+    }
+
+    #[test]
+    fn an_entity_without_the_alias_gets_it() {
+        let seen = Seen {
+            by_alias: None,
+            by_name: Some("e1".to_owned()),
+        };
+        assert_eq!(
+            plan("atlas", &seen),
+            Plan::AddAlias {
+                entity_id: "e1".to_owned()
+            }
+        );
+    }
+
+    #[test]
+    fn an_alias_on_another_entity_while_the_agents_name_is_taken_is_a_conflict() {
+        let seen = Seen {
+            by_alias: Some(entity("e1", "entity_1a2b3c4d.root")),
+            by_name: Some("e2".to_owned()),
+        };
+        assert_eq!(plan("atlas", &seen), conflict("entity_1a2b3c4d.root"));
+    }
+
+    #[test]
+    fn an_alias_on_a_foreign_entity_is_not_moved_to_the_agents() {
+        let seen = Seen {
+            by_alias: Some(entity("e1", "bob")),
+            by_name: Some("e2".to_owned()),
+        };
+        assert_eq!(plan("atlas", &seen), conflict("bob"));
+    }
+}

swarm-controller/src/agent_identity.rs

@@ -8,7 +8,8 @@
 //! and its grant covers exactly the calls made here (`swarm-bao.nix`'s
 //! `controllerPolicyText`: `update` on the agent PKI role's `issue` path, and
 //! `create/update` on `secret/data/swarm/agents/*`, on
-//! `sys/policies/acl/hive-*`, and on `auth/cert/certs/hive-*`).
+//! `sys/policies/acl/hive-*`, and on `auth/cert/certs/hive-*`, plus the
+//! `identity/` and `sys/auth/cert` stanzas [`crate::agent_entity`] uses).
 //!
 //! **Two credentials, deliberately unrelated.** The certificate reaches the
 //! store; the queue secret identifies the agent to the swarm queue. Both sit
@@ -67,13 +68,18 @@ struct RoleInputs<'a> {
     ca: &'a str,
     /// The policy document the role attaches.
     policy: String,
+    /// The policies the role attaches by name: the agent's own, and the one every
+    /// agent shares.
+    attached: [String; 2],
 }
 
 fn role_inputs<'a>(agent: &str, credential: &'a mtls::Credential) -> Result<RoleInputs<'a>> {
+    let name = policy::agent_object_name(agent)?;
     Ok(RoleInputs {
-        name: policy::agent_object_name(agent)?,
+        name: name.clone(),
         ca: &credential.ca,
         policy: policy::render_agent(agent)?,
+        attached: [name, policy::SHARED_AGENT_POLICY.to_owned()],
     })
 }
 
@@ -119,9 +125,9 @@ pub(crate) fn generate_queue_secret() -> Result<String> {
 ///    identity at the swarm queue, minted here so that the credential an agent
 ///    presents names *it* rather than its hive;
 /// 4. write the ACL document [`policy::render_agent`] renders — read and list on
-///    this one agent's paths and nothing else;
-/// 5. write the cert-auth role that ties the three together, pinning the CA
-///    the store named as this leaf's issuer.
+///    this one agent's paths — and the shared [`policy::render_shared_agent`];
+/// 5. write the cert-auth role that ties the three together, attaching both
+///    policies and pinning the CA the store named as this leaf's issuer.
 ///
 /// Policy before role, for the reason `read_policy::provision` gives: the role
 /// names the policy, so the other order leaves a window in which it points at
@@ -199,17 +205,29 @@ pub async fn mint_and_verify(agent: &str) -> Result<()> {
         .write_policy(&inputs.name, &inputs.policy)
         .await
         .with_context(|| format!("writing the read policy {}", inputs.name))?;
+    store
+        .write_policy(policy::SHARED_AGENT_POLICY, &policy::render_shared_agent())
+        .await
+        .with_context(|| {
+            format!(
+                "writing the shared agent policy {}",
+                policy::SHARED_AGENT_POLICY
+            )
+        })?;
     store
         .write_cert_role(
             DEFAULT_CERT_MOUNT,
             &inputs.name,
             inputs.ca,
             &inputs.name,
-            &inputs.name,
+            &inputs.attached.each_ref().map(String::as_str),
         )
         .await
         .with_context(|| format!("writing the cert-auth role {}", inputs.name))?;
     tracing::info!(agent, %path, role = %name, "agent store identity published");
+    // Before the read-back below, whose login would otherwise create the
+    // entity under a generated name.
+    crate::agent_entity::ensure_agent(&store, agent).await;
 
     read_back_as_agent(&credential, &name, &path, &queue_path, &queue_credential).await?;
     tracing::info!(
@@ -450,6 +468,11 @@ mod tests {
             "the CN the leaf is issued for"
         );
         assert_eq!(inputs.ca, "AGENT-CA");
+        assert_eq!(
+            inputs.attached,
+            ["hive-agent-atlas", policy::SHARED_AGENT_POLICY],
+            "the role attaches the agent's own policy and the shared one"
+        );
         assert_eq!(inputs.policy, policy::render_agent("atlas").expect("legal"));
         assert!(!inputs.policy.contains("swarm/hives/"), "{}", inputs.policy);
 

swarm-controller/src/main.rs

@@ -23,6 +23,7 @@ use anyhow::{Context, Result};
 
 mod agent_archetype;
 mod agent_control;
+mod agent_entity;
 mod agent_icon;
 mod agent_identity;
 mod agent_input;

swarm-controller/src/read_policy.rs

@@ -193,7 +193,7 @@ async fn write_role(store: &SecretStore, hive: &str, ca: &str) -> Result<()> {
     store
         // The subject is the hive's name: `glue-bao-tls.nix` mints a hive's
         // client leaf with its name as the CN, and cert auth matches on that.
-        .write_cert_role(DEFAULT_CERT_MOUNT, &name, ca, hive, &name)
+        .write_cert_role(DEFAULT_CERT_MOUNT, &name, ca, hive, &[&name])
         .await
         .with_context(|| format!("writing the cert-auth role {name}"))?;
     tracing::info!(hive, role = %name, "hive store role in place");
@@ -234,9 +234,10 @@ pub async fn ensure_agent_policies() {
     }
 }
 
-/// One pass: log in once, list the agents, write each one's policy.
+/// One pass: log in once, list the agents, write the shared agent policy and
+/// each one's own, and give each its identity entity.
 ///
-/// The policy only. An agent's cert-auth role and certificate are
+/// Policies and entities only. An agent's cert-auth role and certificate are
 /// `agent_identity::mint_and_verify`'s, and that reissues the certificate, so
 /// it is not what a pass on every start may call.
 ///
@@ -260,6 +261,21 @@ async fn rewrite_agent_policies() -> Result<(), Unreachable> {
         .list_cert_roles(DEFAULT_CERT_MOUNT)
         .await
         .map_err(|e| Unreachable(format!("listing the store's cert-auth roles: {e}")))?;
+    // First, so the cert-auth role that names it never points at nothing.
+    match store
+        .write_policy(policy::SHARED_AGENT_POLICY, &policy::render_shared_agent())
+        .await
+    {
+        Ok(()) => tracing::info!(
+            policy = policy::SHARED_AGENT_POLICY,
+            "shared agent policy in place"
+        ),
+        Err(e) => tracing::warn!(
+            policy = policy::SHARED_AGENT_POLICY,
+            error = %e,
+            "writing the shared agent policy failed"
+        ),
+    }
     let policies = agent_policies(&roles);
     let mut written = 0_usize;
     for (name, document) in &policies {
@@ -271,6 +287,7 @@ async fn rewrite_agent_policies() -> Result<(), Unreachable> {
         }
     }
     tracing::info!(written, agents = policies.len(), "agent policies rewritten");
+    crate::agent_entity::ensure_agents(&store, &policy::agents_from_role_names(&roles)).await;
     Ok(())
 }