fix(service): prevent recursive flock deadlock and crontab wipe; increase component RPC timeout

- initd.uc: retry_start_on_wan_up now calls BIN_PATH directly instead of
  SERVICE_INIT to avoid the rc.common flock-1000 self-deadlock that was
  hanging Backend CI for the full 1-hour runner timeout
- initd.uc: schedule_start_retry uses background_command() so inherited
  fd 1000 (procd lock) is closed before the sleep starts
- updates.uc, parental_quota.uc: read /etc/crontabs/root via fs.readfile
  instead of crontab -l — BusyBox exits 1 on an empty file causing
  command_output_from_args to return '' and wipe all user cron entries
- shell/index.ts: COMPONENT_ACTION_RPC_TIMEOUT_MS 10s -> 45s to prevent
  false RPC timeout errors during opkg installs on slow routers

fe-app-tachyon/src/tachyon/methods/shell/index.ts

@@ -11,7 +11,7 @@ const SERVICE_ACTION_TIMEOUT_MS = 2 * 60 * 1000;
 const SERVICE_ACTION_POLL_INTERVAL_MS = 1000;
 const LATENCY_TEST_TIMEOUT_MS = 30 * 1000;
 const LATENCY_TEST_POLL_INTERVAL_MS = 1000;
-const COMPONENT_ACTION_RPC_TIMEOUT_MS = 10000;
+const COMPONENT_ACTION_RPC_TIMEOUT_MS = 45000;
 const COMPONENT_ACTION_POLL_INTERVAL_MS = 1000;
 const COMPONENT_ACTION_SELF_UPDATE_SETTLE_MS = 2000;
 const COMPONENT_ACTION_TRANSIENT_RPC_GRACE_MS = 30000;

luci-app-tachyon/htdocs/luci-static/resources/view/tachyon/main.js

@@ -3157,7 +3157,7 @@ var SERVICE_ACTION_TIMEOUT_MS = 2 * 60 * 1e3;
 var SERVICE_ACTION_POLL_INTERVAL_MS = 1e3;
 var LATENCY_TEST_TIMEOUT_MS = 30 * 1e3;
 var LATENCY_TEST_POLL_INTERVAL_MS = 1e3;
-var COMPONENT_ACTION_RPC_TIMEOUT_MS = 1e4;
+var COMPONENT_ACTION_RPC_TIMEOUT_MS = 45e3;
 var COMPONENT_ACTION_POLL_INTERVAL_MS = 1e3;
 var COMPONENT_ACTION_SELF_UPDATE_SETTLE_MS = 2e3;
 var COMPONENT_ACTION_TRANSIENT_RPC_GRACE_MS = 3e4;

tachyon/files/usr/lib/components/updates.uc

@@ -923,7 +923,11 @@ function log_cron_apply_result(result) {
 }
 
 function remove_cron_jobs(list_marker, subscription_marker, component_marker) {
-    let crontab = command_output_from_args([ "crontab", "-l" ]);
+    // BusyBox `crontab -l` exits with code 1 when the crontab is empty or
+    // absent; command_output_from_args returns "" in that case and a
+    // subsequent write would wipe all user cron lines.  Read the file
+    // directly to avoid this.
+    let crontab = as_string(fs.readfile("/etc/crontabs/root") || "");
     let markers = [ list_marker, subscription_marker, component_marker ];
     let result = {
         crontab: strip_unmarked_tachyon_cron_lines(
@@ -938,10 +942,12 @@ function remove_cron_jobs(list_marker, subscription_marker, component_marker) {
 }
 
 function refresh_cron_from_sources(settings, sections, bin, list_marker, subscription_marker, component_marker) {
+    // Same as remove_cron_jobs: read /etc/crontabs/root directly to avoid
+    // BusyBox's `crontab -l` exit-code-1 on an empty or absent file.
     let result = cron_refresh_apply_result(
         settings,
         sections,
-        command_output_from_args([ "crontab", "-l" ]),
+        as_string(fs.readfile("/etc/crontabs/root") || ""),
         bin,
         list_marker,
         subscription_marker,

tachyon/files/usr/lib/service/initd.uc

@@ -284,10 +284,19 @@ function schedule_start_retry(path, delay_seconds) {
     if (!ensure_parent_dir(path))
         return false;
 
+    // Build the retry worker.  The inner command sleeps, removes the pid file,
+    // then runs the retry action.  We wrap everything in background_command() so
+    // that all inherited descriptors (in particular procd's flock fd 1000) are
+    // closed before the sleep starts — without that the background subshell
+    // would pin the lock file and prevent any subsequent service operation that
+    // also tries to acquire flock 1000.
     let worker = command_from_args([ "sleep", delay_seconds ]) +
         "; " + command_from_args([ "rm", "-f", path ]) +
-        "; exec " + command_from_args([ SERVICE_INIT, "retry_start_on_wan_up" ]);
-    let result = command_capture(command_from_args([ "sh", "-c", worker ]) + " >/dev/null 2>&1 & echo $!");
+        "; " + command_from_args([ SERVICE_INIT, "retry_start_on_wan_up" ]);
+    let bg = background_command(worker);
+    // background_command() already appends " &"; we need $! so run it inside a
+    // subshell that prints the pid back to us.
+    let result = command_capture("{ " + bg + " } ; echo $!");
     let pid = trim(result.output);
     if (result.status != 0 || !numeric_text(pid))
         return false;
@@ -494,7 +503,11 @@ function retry_start_on_wan_up(owner_pid) {
         return 0;
 
     command_success_from_args([ "logger", "-t", SERVICE_NAME, "[info] Retrying failed Tachyon start after WAN came up" ]);
-    return command_status_from_args([ SERVICE_INIT, "restart", "triggered" ]);
+    // Do NOT call SERVICE_INIT here: rc.common holds flock 1000 for the whole
+    // duration of the init.d script, so a nested /etc/init.d/tachyon call would
+    // block forever waiting for that same lock (recursive flock self-deadlock).
+    // Call BIN_PATH directly, the same way start_service and stop_service do.
+    return command_status_from_args([ BIN_PATH, "restart" ]);
 }
 
 function active_service_action_value() {

tachyon/files/usr/lib/service/parental_quota.uc

@@ -347,7 +347,11 @@ function sync_enforcement(blocked_macs, blocked_ips) {
 }
 
 function crontab_lines() {
-    return split(command_output_from_args([ "crontab", "-l" ]), "\n");
+    // BusyBox's `crontab -l` exits with code 1 when the file is empty or
+    // absent, making command_output_from_args() return "" and causing a
+    // subsequent crontab write to discard every user cron line.  Read the
+    // file directly instead; null (no file yet) is treated as empty string.
+    return split(as_string(fs.readfile("/etc/crontabs/root") || ""), "\n");
 }
 
 function cron_line() {