Fixed odd issue with the cli not finding the socket properly

CHANGELOG.md

@@ -4,6 +4,32 @@ All notable changes to this project are documented here. The format
 follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and
 this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [v0.1.1] — 2026-05-15
+
+### Changed
+
+- **`install.sh` now repairs data-dir permissions on every run.**
+  Re-running the installer reasserts the canonical ownership
+  (`quptime:quptime`) and modes across `/etc/quptime/` — `0750` on
+  the dir, `0700` on `keys/`, `0600` on `node.yaml`, `cluster.yaml`,
+  `trust.yaml`, and `keys/private.pem`, `0644` on `keys/public.pem`
+  and `keys/cert.pem`. Makes the installer the one-step recovery
+  path when something has tampered with modes (e.g. a stray
+  `chmod -R`, a backup restore, or an accidental `sudo qu init`
+  that left files owned by root). Unknown files in the dir are left
+  alone.
+
+### Fixed
+
+- **CLI socket lookup as the daemon user.** `sudo -u quptime qu …`
+  no longer fails with `dial daemon socket /tmp/quptime-quptime/…:
+  no such file or directory` while the system daemon is running.
+  `config.SocketPath()` now probes the canonical systemd location
+  (`/run/quptime/quptime.sock`, then `/var/run/quptime/quptime.sock`)
+  regardless of euid before falling back to per-user paths, so the
+  CLI reaches the daemon's socket even when `sudo` has stripped
+  `RUNTIME_DIRECTORY` and `XDG_RUNTIME_DIR` from the environment.
+
 ## [v0.1.0] — 2026-05-15
 
 ### Changed
@@ -112,4 +138,5 @@ Initial public release.
   Planned for a future release.
 
 [v0.0.1]: https://git.cer.sh/axodouble/quptime/releases/tag/v0.0.1
-[v0.1.0]: https://git.cer.sh/axodouble/quptime/releases/tag/v0.1.0
+[v0.1.0]: https://git.cer.sh/axodouble/quptime/releases/tag/v0.1.0
+[v0.1.1]: https://git.cer.sh/axodouble/quptime/releases/tag/v0.1.1

docs/installation.md

@@ -70,6 +70,15 @@ What it does:
    `/etc/systemd/system/quptime.service` (hardened — matches the unit
    in [systemd.md](deployment/systemd.md)). Enables but does not start
    the service, so you can configure identity before first boot.
+5. Repairs ownership and modes under `/etc/quptime/` to the canonical
+   layout (`0750` on the dir, `0700` on `keys/`, `0600` on
+   `node.yaml` / `cluster.yaml` / `trust.yaml` / `keys/private.pem`,
+   `0644` on `keys/public.pem` / `keys/cert.pem`). This makes the
+   installer idempotent for permission damage — if something
+   tightened or loosened modes (a stray `chmod -R`, a misguided
+   backup restore, an accidental `sudo qu init`), re-running
+   `install.sh` puts everything back without touching the contents
+   of those files.
 
 ## Build from source
 

docs/troubleshooting.md

@@ -172,7 +172,9 @@ still see this error, the most likely causes are:
 
 - The data directory is read-only or owned by a different user — the
   bootstrap can't write `node.yaml`. Fix permissions on
-  `$QUPTIME_DIR`.
+  `$QUPTIME_DIR`. The fastest fix on a standard install is just to
+  re-run `install.sh` — it reasserts the canonical ownership and
+  modes on the whole tree without touching your config.
 - Something else removed `node.yaml` mid-run (a config-management
   tool, a misconfigured volume). Re-run `qu serve` and it will
   rebuild from env, or run `qu init` manually with the flags you
@@ -197,7 +199,9 @@ load private key: ...
 ```
 
 Permissions on `keys/private.pem` are wrong — should be 0600 and owned
-by the daemon user. Fix and restart.
+by the daemon user. Fix and restart. Re-running `install.sh` on a
+standard install is the easiest path: it repairs ownership and modes
+on the entire data dir.
 
 ## Probes look much slower than expected
 

install.sh

@@ -175,20 +175,62 @@ fi
 
 install -d -o "$SERVICE_USER" -g "$SERVICE_GROUP" -m 0750 "$DATA_DIR"
 
-# Reassert ownership on the dir's contents. Two cases this catches:
+# Repair ownership and permissions on the data dir's contents. Catches:
 #   - re-running the installer over a previous install where the
-#     service user/group changed
+#     service user/group changed.
 #   - the operator ran `qu init` or `qu serve` as root once (easy
 #     mistake: `sudo qu init` is shorter than the documented
 #     `sudo -u quptime qu init`). When the daemon runs as root its
 #     DataDir() resolves to /etc/quptime, so any files it writes land
-#     here owned by root:root mode 0600 — the systemd service then
-#     fails with `open node.yaml: permission denied`.
-# chown -R only changes ownership, not perms, so file modes set by
-# the daemon (0600 for node.yaml, 0700 for keys/) are preserved.
-if [ -n "$(ls -A "$DATA_DIR" 2>/dev/null)" ]; then
-    chown -R "$SERVICE_USER:$SERVICE_GROUP" "$DATA_DIR"
-fi
+#     owned by root:root — the systemd service then fails with
+#     `open node.yaml: permission denied`.
+#   - someone or something (a stray `chmod -R`, a misguided backup
+#     restore) tightened or loosened modes. Re-running the installer
+#     should be enough to get back to a working baseline.
+# The canonical layout (mirrors the modes the daemon writes itself
+# in internal/config and internal/crypto):
+#   /etc/quptime/                 quptime:quptime  0750
+#   /etc/quptime/keys/            quptime:quptime  0700
+#   /etc/quptime/node.yaml        quptime:quptime  0600
+#   /etc/quptime/cluster.yaml     quptime:quptime  0600
+#   /etc/quptime/trust.yaml       quptime:quptime  0600
+#   /etc/quptime/keys/private.pem quptime:quptime  0600
+#   /etc/quptime/keys/public.pem  quptime:quptime  0644
+#   /etc/quptime/keys/cert.pem    quptime:quptime  0644
+# The runtime dir /var/run/quptime is owned by systemd via
+# RuntimeDirectory= and rebuilt at each service start, so we leave it
+# alone.
+repair_perms() {
+    # Always reset the top-level dir mode — `install -d` only sets it
+    # on creation, not on re-run.
+    chown "$SERVICE_USER:$SERVICE_GROUP" "$DATA_DIR"
+    chmod 0750 "$DATA_DIR"
+
+    # Reassert ownership across the whole tree in one pass.
+    if [ -n "$(ls -A "$DATA_DIR" 2>/dev/null)" ]; then
+        chown -R "$SERVICE_USER:$SERVICE_GROUP" "$DATA_DIR"
+    fi
+
+    # keys/ is a directory with its own tighter mode.
+    if [ -d "$DATA_DIR/keys" ]; then
+        chmod 0700 "$DATA_DIR/keys"
+    fi
+
+    # Each known file gets its canonical mode if it exists. We don't
+    # create anything that isn't already there — that's `qu init`'s
+    # job — and we don't touch unknown files an operator may have
+    # parked in the dir.
+    local f
+    for f in node.yaml cluster.yaml trust.yaml keys/private.pem; do
+        [ -f "$DATA_DIR/$f" ] && chmod 0600 "$DATA_DIR/$f"
+    done
+    for f in keys/public.pem keys/cert.pem; do
+        [ -f "$DATA_DIR/$f" ] && chmod 0644 "$DATA_DIR/$f"
+    done
+}
+
+repair_perms
+echo "> reasserted ownership ($SERVICE_USER:$SERVICE_GROUP) and modes under $DATA_DIR"
 
 echo "> writing $SERVICE_FILE"
 cat > "$SERVICE_FILE" <<'EOF'

internal/config/paths.go

@@ -58,19 +58,20 @@ func DataDir() string {
 // SocketPath returns the unix socket used for local CLI ↔ daemon control.
 //
 // Resolution order:
-//  1. $QUPTIME_SOCKET — explicit operator override
+//  1. $QUPTIME_SOCKET — explicit operator override.
 //  2. $RUNTIME_DIRECTORY — set by systemd when the unit declares
-//     RuntimeDirectory=quptime. This is the path that matters in
-//     practice: with User=quptime + PrivateTmp=true, the daemon's
-//     /tmp is namespaced and invisible to the root CLI shell, so a
-//     /tmp fallback yields "no such file" even though the daemon is
-//     happily listening. Anchoring on $RUNTIME_DIRECTORY puts the
-//     socket at /run/quptime/quptime.sock, which is the same inode
-//     the root-CLI default (/var/run/quptime/…) reaches via the
-//     /var/run → /run symlink.
-//  3. /var/run/quptime/… when euid is 0 (CLI side, packaged installs)
-//  4. $XDG_RUNTIME_DIR/quptime/… for user-mode installs
-//  5. /tmp/quptime-<user>/… as a last resort
+//     RuntimeDirectory=quptime. This is the path the daemon uses
+//     when run under the packaged unit: /run/quptime/quptime.sock.
+//  3. The canonical system socket path — /run/quptime/quptime.sock —
+//     if it exists. This catches the CLI side regardless of who is
+//     invoking it: `sudo -u quptime qu status` strips RUNTIME_DIRECTORY
+//     and XDG_RUNTIME_DIR, so without this probe the CLI falls all
+//     the way through to /tmp/quptime-<user>/… and reports "no such
+//     file" even while the daemon is happily listening.
+//  4. /var/run/quptime/… when euid is 0 (CLI side, packaged installs
+//     on systems where /var/run isn't a symlink to /run).
+//  5. $XDG_RUNTIME_DIR/quptime/… for user-mode installs.
+//  6. /tmp/quptime-<user>/… as a last resort.
 func SocketPath() string {
 	if v := os.Getenv("QUPTIME_SOCKET"); v != "" {
 		return v
@@ -84,6 +85,18 @@ func SocketPath() string {
 		}
 		return filepath.Join(v, SocketName)
 	}
+	// If a system-managed daemon is already listening, route there
+	// regardless of euid. Without this, `sudo -u quptime qu …` can't
+	// find the socket the daemon (also running as quptime) created
+	// via RuntimeDirectory=.
+	for _, p := range []string{
+		"/run/quptime/" + SocketName,
+		"/var/run/quptime/" + SocketName,
+	} {
+		if _, err := os.Stat(p); err == nil {
+			return p
+		}
+	}
 	if os.Geteuid() == 0 {
 		return "/var/run/quptime/" + SocketName
 	}