fix: ensure DHCP service stops when the pool is stopped (#231)

Closes #227

changelog.md

@@ -4,6 +4,7 @@
 
 - IMPROVED: **MTProto Telegram routing is more reliable, especially for media** - the WebSocket routing (bridge mode) now works on phones as well as desktop, loads media more reliably, and always prefers WebSocket regardless of the proxy server's transport setting. Data centers without a WebSocket edge no longer stall when the shared fallback domains get rate-limited. New "Cloudflare Worker" support adds a free, per-user relay (no domain to buy) for accounts where media, reactions, or stickers still fail to load - see Settings -> MTProto Proxy.
 - FIXED: **Copy buttons did nothing when the Web UI was opened over plain HTTP** - copying commands, set exports, or system info silently failed on installs reached by LAN IP (for example `http://192.168.1.1`), because browsers only allow the modern clipboard API over HTTPS or localhost. The fallback copy path now works in those cases too.
+- FIXED: **Memory and CPU slowly climbed when the watchdog repeatedly repaired a site** - each watchdog repair left behind a background task that never stopped, piling up over days until b4 was killed. [#227](https://github.com/DanielLavrushin/b4/issues/227)
 
 ## [1.63.0] - 2026-05-22
 
@@ -12,7 +13,6 @@
 - ADDED: **"Test direct TCP" button for the MTProto proxy** - new button in Settings → MTProto Proxy next to "Test connection". Probes Telegram directly, bypassing the DC Relay. Use it to tell whether a problem is on your relay/VPS or somewhere between b4 and Telegram.
 - ADDED: **Auto-update for geo databases** - new "Auto-update" section in Settings → Geo Databases. Optional "Refresh on startup" re-downloads both files every time b4 starts (handy when files are kept in `/tmp` and get wiped on reboot). A separate "Schedule" picker (Off / Daily / Weekly / Monthly) runs a background refresh on its own. Missing files are also re-downloaded automatically at startup whenever a source URL is set, even with both options off.
 - ADDED: **Per-set MSS Clamping** - new "MSS Clamping" section in the TCP tab of each set, alongside the existing global and per-device options. Lets you set a custom MSS for the connections in one set - for example, a small MSS for a Smart TV streaming YouTube. The set must target by IP, GeoIP, or source device; SNI domain or GeoSite targets alone are not enough.
-
 - FIXED: **Direct TCP path used internal Telegram backend addresses instead of public ones** - b4 was reading addresses from Telegram's `getProxyConfig` (a topology file for MTProxy operators) and dialling them directly. Those backends silently dropped the connection after the handshake because clients are not allowed there. b4 now uses the well-known public data center IPs for direct dialling, like real Telegram clients do.
 - FIXED: **MTProto proxy ignored the DC Relay in Auto mode** - when the upstream transport was set to "Auto" and a DC Relay was configured, b4 still went straight over WebSocket and never used the relay. The relay is now tried first in Auto mode whenever one is configured; WebSocket stays as the fallback. The mode description updates to reflect this.
 - FIXED: **MTProto media data center (DC 203) did not work through the DC Relay** - DC 203 traffic was sent to a port that does not exist on any relay setup, so loading media through the relay failed. DC 203 now reuses the DC 2 relay port (matching the existing WebSocket behaviour), and has a built-in default IP for direct TCP as well.

src/nfq/pool.go

@@ -102,6 +102,10 @@ func (p *Pool) Start() error {
 }
 
 func (p *Pool) Stop() {
+	if p.Dhcp != nil {
+		p.Dhcp.Stop()
+	}
+
 	// Stop the connState cleanup goroutine
 	select {
 	case <-p.stopCleanup: