deploy: a process ceiling on every container (audit I3)

Security audit 2026-10-17, finding I3 (low), stage 4. No compose file set
pids_limit: a fork bomb from a compromised panel or postfix had the host's
pid space as its limit. Now 1024 for SelfPost - the plan said 512, but
Postfix starts up to default_process_limit (100) processes per service,
and a busy server runs three smtpd services, cleanup and smtp at once -
256 for nginx, Caddy, Traefik and rspamd, 64 for certbot. The comment that
said root runs "very briefly" now says which root processes stay.

e2e: checkPidsLimit reads /sys/fs/cgroup/pids.max in the container.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

CHANGELOG.md

@@ -30,6 +30,8 @@ Format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); version
   only a request with no `X-Forwarded-For` from loopback or a private
   address gets the liveness answer - the container's own check, the host,
   a monitor next to it; the internet gets `404`.
+- **Every container has a process ceiling** (audit I3): `pids_limit` 1024
+  for SelfPost, 256 for the proxies and rspamd, 64 for certbot.
 - **The TLS private key no longer has to be readable by Postfix** (audit
   I1): the guide gave it to group 10998 for the panel's backup, and
   `postfix` is in that group, so every Postfix daemon - an smtpd talking to

deploy/antispam/docker-compose.antispam.yml

@@ -32,6 +32,8 @@ services:
     # (its Dockerfile, USER 11333:11333).
     image: rspamd/rspamd:3.11.1-3
     restart: unless-stopped
+    # A ceiling on processes and threads (security audit 2026-10-17, I3).
+    pids_limit: 256
     # No host ports: only SelfPost talks to the milter on the compose network.
     volumes:
       # The verdict headers SelfPost's inbound journal reads.

deploy/caddy/docker-compose.caddy.yml

@@ -48,6 +48,8 @@ services:
   caddy:
     image: caddy:2
     restart: unless-stopped
+    # A ceiling on processes and threads (security audit 2026-10-17, I3).
+    pids_limit: 256
     depends_on:
       - selfpost
     ports:

deploy/docker-compose.yml

@@ -99,10 +99,12 @@ services:
       # when the variable is true.
       - "25:25"
       - "127.0.0.1:8080:8080"
-    # Hardening. SelfPost's entrypoint still needs to run as
+    # Hardening. SelfPost's entrypoint runs as root to fix /data ownership
-    # root very briefly to fix /data ownership and normalise permissions
+    # and normalise permissions under the shared `selfpost` group (see
-    # under the shared `selfpost` group (see build/entrypoint.sh) before
+    # build/entrypoint.sh) before supervisord drops the panel to an
-    # supervisord drops the panel to an unprivileged user — so this cannot be
+    # unprivileged user; supervisord, Postfix's master, cert-reload and
+    # logrotate stay root for the container's life, which is inherent to
+    # them (security audit 2026-10-17, I3) — so this cannot be
     # `user: panel` or a fully read-only rootfs without breaking that startup
     # self-healing. What IS applied: no privilege escalation past what the
     # image already grants, and every Linux capability dropped except the
@@ -151,3 +153,9 @@ services:
       - SETGID
       - DAC_OVERRIDE
       - KILL
+    # A ceiling on processes and threads, so that a runaway or compromised
+    # program cannot use up the host's (security audit 2026-10-17, I3).
+    # Postfix starts up to 100 processes per service by default
+    # (default_process_limit), and a busy server runs three smtpd services,
+    # cleanup and smtp at once; 1024 leaves them room.
+    pids_limit: 1024

deploy/nginx/docker-compose.nginx.yml

@@ -47,6 +47,8 @@ services:
   nginx:
     image: nginx:1.27
     restart: unless-stopped
+    # A ceiling on processes and threads (security audit 2026-10-17, I3).
+    pids_limit: 256
     depends_on:
       - selfpost
     ports:
@@ -59,6 +61,7 @@ services:
 
   certbot:
     image: certbot/certbot:latest
+    pids_limit: 64
     volumes:
       - ./nginx/certbot-etc:/etc/letsencrypt
       - ./nginx/certbot-www:/var/www/certbot

deploy/traefik/docker-compose.traefik.yml

@@ -40,6 +40,8 @@ services:
   traefik:
     image: traefik:v3.1
     restart: unless-stopped
+    # A ceiling on processes and threads (security audit 2026-10-17, I3).
+    pids_limit: 256
     depends_on:
       - selfpost
     command:

docs/security-audit-2026-10-17.md

@@ -799,8 +799,11 @@ the argument.
 
 ### Stage 4 — resources and imports (~5 h)
 
-- [ ] **I3.** `pids_limit: 512` in every compose fragment; the comment
+- [x] **I3.** `pids_limit: 512` in every compose fragment; the comment
-  corrected; one e2e run. — **Sonnet**, 0.5 h.
+  corrected; one e2e run. — **Sonnet**, 0.5 h. *Done 2026-10-17:* 1024 for
+  SelfPost (Postfix's default_process_limit of 100 per service, three smtpd
+  services), 256 for the proxies and rspamd, 64 for certbot; e2e reads
+  `pids.max` back (checkPidsLimit).
 - [ ] **W6/M9.** Coalesce Postfix reloads (reload only when the rendered map
   changed; a 2 s debounce for bursts); a per-user limiter on `dns-recheck`.
   — **Opus**, 1.5 h.

test/e2e/main_test.go

@@ -111,6 +111,9 @@ func TestE2E(t *testing.T) {
 		if err := checkPostfixHardening(h); err != nil {
 			t.Fatal(err)
 		}
+		if err := checkPidsLimit(h); err != nil {
+			t.Fatal(err)
+		}
 		if err := checkInboundRelayOff(h); err != nil {
 			t.Fatal(err)
 		}

test/e2e/process_check.go

@@ -146,6 +146,19 @@ func checkDKIMTree(s *stack) error {
 	return nil
 }
 
+// checkPidsLimit reads the container's process ceiling back from the kernel:
+// the shipped compose sets pids_limit (security audit 2026-10-17, I3).
+func checkPidsLimit(s *stack) error {
+	out, err := s.execIn("selfpost", "cat", "/sys/fs/cgroup/pids.max")
+	if err != nil {
+		return fmt.Errorf("read pids.max: %v\n%s", err, out)
+	}
+	if got := strings.TrimSpace(out); got != "1024" {
+		return fmt.Errorf("the container's pids.max is %q, want the compose file's 1024", got)
+	}
+	return nil
+}
+
 // checkPostfixHardening reads back what postfix-config.sh sets against
 // downgrade and enumeration (audit 2026-10-12, L31).
 func checkPostfixHardening(s *stack) error {