fix(#702): route container journal reads through hive-priv
The privsep drop to the hive-core user left four journalctl -M <container> call sites shelling out directly. -M enters the container namespace via the machine bus, which needs root, so all container-journal reads failed with Permission denied. Add a ReadContainerJournal verb to hive-priv and route dashboard get_journal, manager get_logs, the rebuild-failure journal tail, and the agent host-journal -M path through it. Host-journal reads (no -M) stay direct via systemd-journal group membership.
This commit is contained in:
parent
5da7f6cd3a
commit
9e12012a95
7 changed files with 279 additions and 65 deletions
|
|
@ -378,6 +378,36 @@ pub async fn dispatch_host_journal(
|
|||
};
|
||||
}
|
||||
let n = lines.unwrap_or(30).min(100);
|
||||
|
||||
// A container (`-M`) read enters the container namespace and needs
|
||||
// root, so it's delegated to hive-priv. A host read (no container)
|
||||
// the unprivileged hive-core user can do directly via its
|
||||
// systemd-journal group membership.
|
||||
if let Some(c) = container {
|
||||
tracing::info!(%agent, machine = %c, %n, "get_host_journal (container)");
|
||||
return match crate::priv_client::read_container_journal(
|
||||
c,
|
||||
n,
|
||||
false,
|
||||
hive_sh4re::priv_proto::JournalOutput::Short,
|
||||
unit.clone(),
|
||||
priority.as_ref().map(|p| p.as_str().to_owned()),
|
||||
grep.clone(),
|
||||
since.clone(),
|
||||
until.clone(),
|
||||
)
|
||||
.await
|
||||
{
|
||||
Ok((stdout, stderr)) => {
|
||||
let content = if !stdout.is_empty() { stdout } else { stderr };
|
||||
AgentResponse::HostJournal { content }
|
||||
}
|
||||
Err(e) => AgentResponse::Err {
|
||||
message: format!("journal read: {e:#}"),
|
||||
},
|
||||
};
|
||||
}
|
||||
|
||||
let mut args: Vec<String> = vec![
|
||||
"--no-pager".to_owned(),
|
||||
"--output=short".to_owned(),
|
||||
|
|
@ -388,10 +418,6 @@ pub async fn dispatch_host_journal(
|
|||
args.push("-u".to_owned());
|
||||
args.push(u.clone());
|
||||
}
|
||||
if let Some(c) = container {
|
||||
args.push("-M".to_owned());
|
||||
args.push(c.clone());
|
||||
}
|
||||
if let Some(p) = priority {
|
||||
args.push("-p".to_owned());
|
||||
args.push(p.as_str().to_owned());
|
||||
|
|
|
|||
|
|
@ -1160,10 +1160,10 @@ struct JournalQuery {
|
|||
lines: Option<u32>,
|
||||
}
|
||||
|
||||
/// Shell out to `journalctl -M <container> -b` and return its text
|
||||
/// output. Operator-only by virtue of the dashboard being host-bound;
|
||||
/// hive-c0re already runs as root in its systemd unit so journalctl
|
||||
/// has the access it needs.
|
||||
/// Read `journalctl -M <container> -b` and return its text output.
|
||||
/// Operator-only by virtue of the dashboard being host-bound. hive-c0re
|
||||
/// runs unprivileged (privsep), so the `-M` read — which enters the
|
||||
/// container namespace and needs root — is delegated to hive-priv.
|
||||
async fn get_journal(
|
||||
AxumPath(name): AxumPath<String>,
|
||||
axum::extract::Query(q): axum::extract::Query<JournalQuery>,
|
||||
|
|
@ -1184,40 +1184,45 @@ async fn get_journal(
|
|||
return error_response(&format!("journal: no managed container {prefixed:?}"));
|
||||
}
|
||||
let lines = q.lines.unwrap_or(500).min(5000);
|
||||
let mut cmd = tokio::process::Command::new("journalctl");
|
||||
cmd.args([
|
||||
"-M",
|
||||
&prefixed,
|
||||
"-b",
|
||||
"--no-pager",
|
||||
"--output=short-iso",
|
||||
"--lines",
|
||||
])
|
||||
.arg(lines.to_string());
|
||||
if let Some(u) = q.unit.as_deref().filter(|s| !s.is_empty()) {
|
||||
// accept hive-ag3nt[.service] — anything else refused.
|
||||
let allowed = ["hive-ag3nt.service"];
|
||||
let unit = if u.ends_with(".service") {
|
||||
u.to_owned()
|
||||
} else {
|
||||
format!("{u}.service")
|
||||
};
|
||||
if !allowed.contains(&unit.as_str()) {
|
||||
return error_response(&format!("journal: unknown unit {unit:?}"));
|
||||
let unit = match q.unit.as_deref().filter(|s| !s.is_empty()) {
|
||||
Some(u) => {
|
||||
// accept hive-ag3nt[.service] — anything else refused.
|
||||
let allowed = ["hive-ag3nt.service"];
|
||||
let unit = if u.ends_with(".service") {
|
||||
u.to_owned()
|
||||
} else {
|
||||
format!("{u}.service")
|
||||
};
|
||||
if !allowed.contains(&unit.as_str()) {
|
||||
return error_response(&format!("journal: unknown unit {unit:?}"));
|
||||
}
|
||||
Some(unit)
|
||||
}
|
||||
cmd.args(["-u", &unit]);
|
||||
}
|
||||
match cmd.output().await {
|
||||
Ok(out) => {
|
||||
None => None,
|
||||
};
|
||||
match crate::priv_client::read_container_journal(
|
||||
&prefixed,
|
||||
lines,
|
||||
true,
|
||||
hive_sh4re::priv_proto::JournalOutput::ShortIso,
|
||||
unit,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
)
|
||||
.await
|
||||
{
|
||||
Ok((stdout, stderr)) => {
|
||||
// Combine stdout + stderr — journalctl emits to both on errors.
|
||||
let mut body = String::from_utf8_lossy(&out.stdout).into_owned();
|
||||
if !out.status.success() {
|
||||
let mut body = stdout;
|
||||
if !stderr.is_empty() {
|
||||
body.push_str("\n--- stderr ---\n");
|
||||
body.push_str(&String::from_utf8_lossy(&out.stderr));
|
||||
body.push_str(&stderr);
|
||||
}
|
||||
([("content-type", "text/plain; charset=utf-8")], body).into_response()
|
||||
}
|
||||
Err(e) => error_response(&format!("journalctl spawn: {e}")),
|
||||
Err(e) => error_response(&format!("journal read: {e:#}")),
|
||||
}
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -1195,14 +1195,24 @@ async fn priv_run(kind: &str, name: &str) -> Result<()> {
|
|||
/// or when the journal can't be read (machine gone, journalctl
|
||||
/// missing); it never produces an error of its own.
|
||||
async fn container_journal_tail(container: &str) -> String {
|
||||
let out = Command::new("journalctl")
|
||||
.args(["-M", container, "-n", "40", "--no-pager", "--output=short"])
|
||||
.output()
|
||||
.await;
|
||||
match out {
|
||||
Ok(o) if !o.stdout.is_empty() => format!(
|
||||
// `-M` enters the container namespace and needs root, so the read
|
||||
// is delegated to hive-priv (hive-c0re itself runs unprivileged).
|
||||
let res = crate::priv_client::read_container_journal(
|
||||
container,
|
||||
40,
|
||||
false,
|
||||
hive_sh4re::priv_proto::JournalOutput::Short,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
)
|
||||
.await;
|
||||
match res {
|
||||
Ok((stdout, _)) if !stdout.is_empty() => format!(
|
||||
"\n--- last 40 journal lines from container '{container}' ---\n{}",
|
||||
String::from_utf8_lossy(&o.stdout).trim_end()
|
||||
stdout.trim_end()
|
||||
),
|
||||
_ => String::new(),
|
||||
}
|
||||
|
|
|
|||
|
|
@ -254,31 +254,28 @@ async fn dispatch(req: &ManagerRequest, coord: &Arc<Coordinator>) -> ManagerResp
|
|||
let n = lines.unwrap_or(50);
|
||||
// `journalctl -M` wants the container name (`h-<name>`),
|
||||
// not the logical agent name. `container_name` adds the prefix.
|
||||
// The `-M` read needs root, so it goes through hive-priv.
|
||||
let machine = crate::lifecycle::container_name(agent);
|
||||
tracing::info!(%agent, %machine, %n, "manager: get_logs");
|
||||
match tokio::process::Command::new("journalctl")
|
||||
.args([
|
||||
"-M",
|
||||
&machine,
|
||||
"-n",
|
||||
&n.to_string(),
|
||||
"--no-pager",
|
||||
"--output=short",
|
||||
])
|
||||
.output()
|
||||
.await
|
||||
match crate::priv_client::read_container_journal(
|
||||
&machine,
|
||||
n,
|
||||
false,
|
||||
hive_sh4re::priv_proto::JournalOutput::Short,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
None,
|
||||
)
|
||||
.await
|
||||
{
|
||||
Ok(out) => {
|
||||
let content = if out.status.success() || !out.stdout.is_empty() {
|
||||
String::from_utf8_lossy(&out.stdout).into_owned()
|
||||
} else {
|
||||
let stderr = String::from_utf8_lossy(&out.stderr);
|
||||
format!("journalctl exited {}: {stderr}", out.status)
|
||||
};
|
||||
Ok((stdout, stderr)) => {
|
||||
let content = if !stdout.is_empty() { stdout } else { stderr };
|
||||
ManagerResponse::Logs { content }
|
||||
}
|
||||
Err(e) => ManagerResponse::Err {
|
||||
message: format!("journalctl spawn failed: {e:#}"),
|
||||
message: format!("get_logs: {e:#}"),
|
||||
},
|
||||
}
|
||||
}
|
||||
|
|
|
|||
|
|
@ -7,7 +7,7 @@
|
|||
//! a persistent connection.
|
||||
|
||||
use anyhow::{Context as _, Result, bail};
|
||||
use hive_sh4re::priv_proto::{BindMount, PRIV_SOCK, PrivRequest, PrivResponse};
|
||||
use hive_sh4re::priv_proto::{BindMount, JournalOutput, PRIV_SOCK, PrivRequest, PrivResponse};
|
||||
use tokio::io::{AsyncBufReadExt, AsyncWriteExt, BufReader};
|
||||
use tokio::net::UnixStream;
|
||||
|
||||
|
|
@ -81,6 +81,37 @@ pub async fn list_containers() -> Result<String> {
|
|||
Ok(stdout)
|
||||
}
|
||||
|
||||
/// Read a container's journal via the root helper (`journalctl -M`).
|
||||
/// Returns `(stdout, stderr)`; a non-zero journalctl exit is reported in
|
||||
/// `stderr` rather than as an `Err`, so callers can surface either.
|
||||
#[allow(clippy::too_many_arguments)]
|
||||
pub async fn read_container_journal(
|
||||
container: &str,
|
||||
lines: u32,
|
||||
boot: bool,
|
||||
output: JournalOutput,
|
||||
unit: Option<String>,
|
||||
priority: Option<String>,
|
||||
grep: Option<String>,
|
||||
since: Option<String>,
|
||||
until: Option<String>,
|
||||
) -> Result<(String, String)> {
|
||||
check(
|
||||
call(&PrivRequest::ReadContainerJournal {
|
||||
container: container.to_owned(),
|
||||
lines,
|
||||
boot,
|
||||
output,
|
||||
unit,
|
||||
priority,
|
||||
grep,
|
||||
since,
|
||||
until,
|
||||
})
|
||||
.await?,
|
||||
)
|
||||
}
|
||||
|
||||
pub async fn write_nspawn_flags(container: &str, binds: &[BindMount]) -> Result<()> {
|
||||
ok(call(&PrivRequest::WriteNspawnFlags {
|
||||
container: container.to_owned(),
|
||||
|
|
|
|||
|
|
@ -21,8 +21,8 @@ use std::path::{Path, PathBuf};
|
|||
|
||||
use anyhow::{Context as _, Result, bail};
|
||||
use hive_sh4re::priv_proto::{
|
||||
AGENT_PREFIX, BindMount, MANAGER_NAME, META_DIR, PRIV_SOCK, PrivRequest, PrivResponse,
|
||||
SIBLING_CONTAINERS,
|
||||
AGENT_PREFIX, BindMount, JournalOutput, MANAGER_NAME, META_DIR, PRIV_SOCK, PrivRequest,
|
||||
PrivResponse, SIBLING_CONTAINERS,
|
||||
};
|
||||
use tokio::io::{AsyncBufReadExt, AsyncWriteExt, BufReader};
|
||||
use tokio::net::{UnixListener, UnixStream};
|
||||
|
|
@ -186,6 +186,24 @@ async fn exec(req: PrivRequest) -> Result<(String, String)> {
|
|||
|
||||
PrivRequest::ListContainers => container_run(&["list"]).await,
|
||||
|
||||
PrivRequest::ReadContainerJournal {
|
||||
ref container,
|
||||
lines,
|
||||
boot,
|
||||
output,
|
||||
ref unit,
|
||||
ref priority,
|
||||
ref grep,
|
||||
ref since,
|
||||
ref until,
|
||||
} => {
|
||||
validate_container_system_name(container)?;
|
||||
read_container_journal(
|
||||
container, lines, boot, output, unit, priority, grep, since, until,
|
||||
)
|
||||
.await
|
||||
}
|
||||
|
||||
PrivRequest::WriteNspawnFlags {
|
||||
ref container,
|
||||
ref binds,
|
||||
|
|
@ -313,6 +331,71 @@ async fn container_run(args: &[&str]) -> Result<(String, String)> {
|
|||
Ok((stdout, stderr))
|
||||
}
|
||||
|
||||
/// Read a container's journal as root via `journalctl -M`. Returns
|
||||
/// `(stdout, stderr)`. Unlike `container_run` a non-zero exit is *not* a
|
||||
/// hard error — journalctl's own diagnostic (folded into `stderr` with
|
||||
/// the exit status) is what the caller surfaces to the operator, so the
|
||||
/// helper never bails.
|
||||
#[allow(clippy::too_many_arguments)]
|
||||
async fn read_container_journal(
|
||||
container: &str,
|
||||
lines: u32,
|
||||
boot: bool,
|
||||
output: JournalOutput,
|
||||
unit: &Option<String>,
|
||||
priority: &Option<String>,
|
||||
grep: &Option<String>,
|
||||
since: &Option<String>,
|
||||
until: &Option<String>,
|
||||
) -> Result<(String, String)> {
|
||||
let mut args: Vec<String> = vec![
|
||||
"-M".to_owned(),
|
||||
container.to_owned(),
|
||||
"--no-pager".to_owned(),
|
||||
format!("--output={}", output.as_journalctl()),
|
||||
"-n".to_owned(),
|
||||
lines.to_string(),
|
||||
];
|
||||
if boot {
|
||||
args.push("-b".to_owned());
|
||||
}
|
||||
if let Some(u) = unit {
|
||||
args.push("-u".to_owned());
|
||||
args.push(u.clone());
|
||||
}
|
||||
if let Some(p) = priority {
|
||||
args.push("-p".to_owned());
|
||||
args.push(p.clone());
|
||||
}
|
||||
// `--grep=`/`--since=`/`--until=` use the `=`-joined form so a value
|
||||
// can never be parsed as a separate journalctl flag.
|
||||
if let Some(g) = grep {
|
||||
args.push(format!("--grep={g}"));
|
||||
}
|
||||
if let Some(s) = since {
|
||||
args.push(format!("--since={s}"));
|
||||
}
|
||||
if let Some(u) = until {
|
||||
args.push(format!("--until={u}"));
|
||||
}
|
||||
let out = Command::new("journalctl")
|
||||
.args(&args)
|
||||
.output()
|
||||
.await
|
||||
.context("invoke journalctl -M")?;
|
||||
let stdout = String::from_utf8_lossy(&out.stdout).into_owned();
|
||||
let stderr = if out.status.success() {
|
||||
String::from_utf8_lossy(&out.stderr).into_owned()
|
||||
} else {
|
||||
format!(
|
||||
"journalctl -M {container} exited {}: {}",
|
||||
out.status,
|
||||
String::from_utf8_lossy(&out.stderr).trim()
|
||||
)
|
||||
};
|
||||
Ok((stdout, stderr))
|
||||
}
|
||||
|
||||
/// Return the system container name for a logical agent name.
|
||||
/// All agents (including the manager) use the `h-` prefix.
|
||||
fn container_system_name(name: &str) -> String {
|
||||
|
|
|
|||
|
|
@ -22,6 +22,29 @@ pub const SIBLING_CONTAINERS: &[&str] = &["hive-forge", "hive-matrix", "hive-gat
|
|||
/// `{META_DIR}#{name}`, derived by `hive-priv` — never passed over the wire.
|
||||
pub const META_DIR: &str = "/var/lib/hyperhive/meta";
|
||||
|
||||
/// Output format for `ReadContainerJournal`. Maps to journalctl
|
||||
/// `--output=<...>`. Restricted to the two formats hive callers use so
|
||||
/// the wire type can't smuggle an arbitrary `--output` value.
|
||||
#[derive(Debug, Clone, Copy, Default, Serialize, Deserialize)]
|
||||
#[serde(rename_all = "snake_case")]
|
||||
pub enum JournalOutput {
|
||||
/// `short` — the journalctl default (syslog-style timestamps).
|
||||
#[default]
|
||||
Short,
|
||||
/// `short-iso` — ISO 8601 timestamps.
|
||||
ShortIso,
|
||||
}
|
||||
|
||||
impl JournalOutput {
|
||||
/// The string journalctl expects after `--output=`.
|
||||
pub fn as_journalctl(self) -> &'static str {
|
||||
match self {
|
||||
JournalOutput::Short => "short",
|
||||
JournalOutput::ShortIso => "short-iso",
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// One bind-mount entry for `WriteNspawnFlags`.
|
||||
/// hive-priv constructs `--bind=<host_path>:<container_path>` (or `--bind-ro=`)
|
||||
/// and validates both paths before writing the conf file.
|
||||
|
|
@ -65,6 +88,45 @@ pub enum PrivRequest {
|
|||
/// `nixos-container list`
|
||||
ListContainers,
|
||||
|
||||
// --- Container journal reads ---
|
||||
/// Read a container's journal via `journalctl -M <container>`.
|
||||
/// Requires root: the machine-bus transport enters the container's
|
||||
/// namespace, so this can't run from the unprivileged hive-c0re
|
||||
/// process. hive-priv validates `container` against the managed-
|
||||
/// container allowlist, then runs journalctl and returns its output.
|
||||
///
|
||||
/// The filters (`unit` / `priority` / `grep` / `since` / `until`)
|
||||
/// are applied within the already-authorized machine and passed to
|
||||
/// journalctl as plain argument values; they can't widen access
|
||||
/// beyond the validated `container`.
|
||||
ReadContainerJournal {
|
||||
/// System container name (`h-<agent>` or a sibling service).
|
||||
container: String,
|
||||
/// `-n <lines>`.
|
||||
lines: u32,
|
||||
/// `-b` — restrict to the current boot.
|
||||
#[serde(default)]
|
||||
boot: bool,
|
||||
/// `--output=<...>`.
|
||||
#[serde(default)]
|
||||
output: JournalOutput,
|
||||
/// `-u <unit>`.
|
||||
#[serde(default)]
|
||||
unit: Option<String>,
|
||||
/// `-p <priority>`.
|
||||
#[serde(default)]
|
||||
priority: Option<String>,
|
||||
/// `--grep=<regex>`.
|
||||
#[serde(default)]
|
||||
grep: Option<String>,
|
||||
/// `--since=<ts>`.
|
||||
#[serde(default)]
|
||||
since: Option<String>,
|
||||
/// `--until=<ts>`.
|
||||
#[serde(default)]
|
||||
until: Option<String>,
|
||||
},
|
||||
|
||||
// --- Config file writes ---
|
||||
/// Update `/etc/nixos-containers/<container>.conf`: strip network-isolation
|
||||
/// vars, force `PRIVATE_NETWORK=0`, and set `EXTRA_NSPAWN_FLAGS` from the
|
||||
|
|
|
|||
Loading…
Reference in a new issue