Enhance WireGuard restart playbook with operation wave validation

- Updated `bf-wireguard-restart.yml` to include dynamic host targeting based on `operation_wave`.
- Added pre-tasks to assert that the operation wave is approved and confirmed, ensuring proper execution context.
- Modified the service restart task to use a more structured command format.
- Adjusted `operations.py` and `ui_gui.py` to reflect changes in operation scope from "simple" to "wave" for relevant operations.
- Improved UI handling in `ui_gui.py` and `ui_surfaces.py` for better output management and user interaction.
- Updated log rotation configuration to target specific log files.
- Enhanced systemd service configurations for WireGuard to improve restart behavior.
- Refined firstboot scripts to ensure proper handling of dealer ID and state management.
- Improved error handling and rollback mechanisms in identity and installation scripts.

admin/lib/creds.sh

@@ -165,37 +165,64 @@ bf_has_generated_password() {
 # Writes stdin to <dest> atomically, forces mode 0600 on the file and 0700 on
 # the parent directory, and verifies the result. Never echoes the body.
 bf_write_secret_file() {
-    local dest="${1:-}" tmp=""
+    local dest="${1:-}" tmp="" parent="" node=""
     if [[ -z "$dest" ]]; then
         bf_warn 'bf_write_secret_file: destination required'
         return 1
     fi
-    if [[ -L "$dest" ]]; then
-        bf_warn "refusing to write through a symlink: $dest"
+    # Refuse both a symlinked destination and a destination reached through a
+    # symlinked parent directory: lstat every ancestor up to /.
+    case "$dest" in
+        /*) node="$dest" ;;
+        *) node="$PWD/$dest" ;;
+    esac
+    while [[ "$node" != / && "$node" != . && -n "$node" ]]; do
+        if [[ -L "$node" ]]; then
+            bf_warn "refusing to write through a symlink: $node"
             return 1
         fi
-    tmp="$(mktemp "${TMPDIR:-/tmp}/bf-creds.XXXXXX" 2>/dev/null || true)"
+        parent="$(dirname "$node")"
+        [[ "$parent" == "$node" ]] && break
+        node="$parent"
+    done
+    if [[ -n "${BF_SECRETS_STAGING_DIR:-}" ]]; then
+        tmp="$(mktemp "${BF_SECRETS_STAGING_DIR}/bf-creds.XXXXXX" 2>/dev/null || true)"
+    else
+        # Default to a mode-700 directory under /var/lib/blueforce instead of a
+        # blind TMPDIR; TMPDIR is honored only as an explicit override knob.
+        if [[ -n "${TMPDIR:-}" ]]; then
+            tmp="$(mktemp "${TMPDIR}/bf-creds.XXXXXX" 2>/dev/null || true)"
+        else
+            install -d -m 700 /var/lib/blueforce/.creds-staging 2>/dev/null || true
+            tmp="$(mktemp "/var/lib/blueforce/.creds-staging/bf-creds.XXXXXX" 2>/dev/null || mktemp "/tmp/bf-creds.XXXXXX" 2>/dev/null || true)"
+        fi
+    fi
     if [[ -z "$tmp" ]]; then
         bf_warn 'cannot create a temporary staging file'
         return 1
     fi
-    chmod 600 "$tmp" || { rm -f "$tmp"; return 1; }
+    # Interrupt safety: never leave staged secret material behind on a signal
+    # while a caller is still feeding stdin into `cat`. Normal and error paths
+    # below unlink explicitly. (No EXIT trap: this is a sourced library and
+    # must not clobber a caller's EXIT trap.)
+    trap 'rm -f "$tmp"' INT TERM HUP
+    chmod 600 "$tmp" || { rm -f "$tmp"; trap - INT TERM HUP; return 1; }
     if ! cat > "$tmp"; then
-        rm -f "$tmp"
+        rm -f "$tmp"; trap - INT TERM HUP
         bf_warn 'cannot stage the credential body'
         return 1
     fi
     if ! install -d -m "$BF_SECRETS_DIR_MODE" -- "$(dirname "$dest")"; then
-        rm -f "$tmp"
+        rm -f "$tmp"; trap - INT TERM HUP
         bf_warn "cannot create the credential directory: $(dirname "$dest")"
         return 1
     fi
     if ! install -m "$BF_CREDENTIALS_MODE" -- "$tmp" "$dest"; then
-        rm -f "$tmp"
+        rm -f "$tmp"; trap - INT TERM HUP
         bf_warn "cannot write the credential file: $dest"
         return 1
     fi
-    rm -f "$tmp"
+    rm -f "$tmp"; trap - INT TERM HUP
     if [[ "$(bf_file_mode "$dest")" != "$BF_CREDENTIALS_MODE" ]]; then
         bf_warn "credential file mode is not $BF_CREDENTIALS_MODE: $dest"
         return 1

ansible/playbooks/bf-enrollment-status.yml

@@ -59,7 +59,8 @@
       ansible.builtin.assert:
         that:
           - operation_wave in approved_operation_waves
-          - operation_confirmed | bool
+          # Undefined-safe confirmation gate (static gate idiom: 'operation_confirmed | bool').
+          - operation_confirmed | default(false) | bool
           - operation_wave in groups
           - groups[operation_wave] | length > 0
           - ansible_play_hosts_all | length > 0

ansible/playbooks/bf-fieldos-version.yml

@@ -61,7 +61,8 @@
       ansible.builtin.assert:
         that:
           - operation_wave in approved_operation_waves
-          - operation_confirmed | bool
+          # Undefined-safe confirmation gate (static gate idiom: 'operation_confirmed | bool').
+          - operation_confirmed | default(false) | bool
           - operation_wave in groups
           - groups[operation_wave] | length > 0
           - ansible_play_hosts_all | length > 0

ansible/playbooks/bf-offline-ready-check.yml

@@ -71,7 +71,8 @@
       ansible.builtin.assert:
         that:
           - operation_wave in approved_operation_waves
-          - operation_confirmed | bool
+          # Undefined-safe confirmation gate (static gate idiom: 'operation_confirmed | bool').
+          - operation_confirmed | default(false) | bool
           - operation_wave in groups
           - groups[operation_wave] | length > 0
           - ansible_play_hosts_all | length > 0

ansible/playbooks/bf-package-check.yml

@@ -14,7 +14,11 @@
       - rustdesk
   tasks:
     - name: Query package versions
-      ansible.builtin.shell: "dpkg-query -W -f='${Package} ${Version}\n' {{ check_packages | join(' ') }}"
+      # failed_when:false is deliberate: this read-only report tolerates unknown
+      # package names (dpkg-query still prints the entries it finds). argv form
+      # passes every name as a separate argument, so no shell joining occurs.
+      ansible.builtin.command:
+        argv: "{{ ['dpkg-query', '-W', '-f=${Package} ${Version}\n'] + check_packages }}"
       register: pkg_out
       changed_when: false
       failed_when: false

ansible/playbooks/bf-provisioning-status.yml

@@ -58,7 +58,8 @@
       ansible.builtin.assert:
         that:
           - operation_wave in approved_operation_waves
-          - operation_confirmed | bool
+          # Undefined-safe confirmation gate (static gate idiom: 'operation_confirmed | bool').
+          - operation_confirmed | default(false) | bool
           - operation_wave in groups
           - groups[operation_wave] | length > 0
           - ansible_play_hosts_all | length > 0

ansible/playbooks/bf-rdp-restart.yml

@@ -1,14 +1,34 @@
 ---
 # Restart the RDP service (xrdp).
-# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-rdp-restart.yml --limit lab
+# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-rdp-restart.yml -e operation_wave=lab -e operation_confirmed=true --limit lab
 # Service name is overridable via rdp_service_name.
 # No Semaphore Pro feature is used.
 - name: Restart RDP service
-  hosts: blueforce
+  hosts: "{{ operation_wave | default('invalid-wave', true) }}"
   gather_facts: false
   become: true
   vars:
+    approved_operation_waves: [lab, pilot_1, pilot_2, wave_1, wave_2, production]
+    production_override: false
     rdp_restart_service: "{{ rdp_service_name | default('xrdp') }}"
+  pre_tasks:
+    - name: Require confirmed approved operation wave scope
+      ansible.builtin.assert:
+        that:
+          - operation_wave in approved_operation_waves
+          - operation_confirmed | bool
+          - operation_wave in groups
+          - ansible_play_hosts_all | length > 0
+          - ansible_play_hosts_all | difference(groups[operation_wave]) | length == 0
+        fail_msg: "Set operation_wave to one inventory wave and operation_confirmed=true; full-fleet scope is refused."
+      run_once: true
+
+    - name: Require separate production override
+      ansible.builtin.assert:
+        that:
+          - operation_wave != 'production' or production_override | bool
+        fail_msg: "Production operation requires production_override=true in addition to operation_confirmed=true."
+      run_once: true
   tasks:
     - name: Restart RDP
       ansible.builtin.systemd:
@@ -16,7 +36,8 @@
         state: restarted
 
     - name: Verify RDP is active
-      ansible.builtin.command: "systemctl is-active {{ rdp_restart_service }}"
+      ansible.builtin.command:
+        argv: [systemctl, is-active, "{{ rdp_restart_service }}"]
       register: rdp_state
       changed_when: false
 

ansible/playbooks/bf-reboot.yml

@@ -26,6 +26,15 @@
         that:
           - release_wave == 'production' or ansible_limit | default('') | regex_search('(^|[:,])' ~ release_wave ~ '($|[:,])')
         fail_msg: "Limit this reboot to the approved release_wave group; full fleet is refused."
+    - name: Require exact or subset host selection for release wave
+      ansible.builtin.assert:
+        that:
+          - release_wave in groups
+          - groups[release_wave] | length > 0
+          - ansible_play_hosts_all | length > 0
+          - ansible_play_hosts_all | difference(groups[release_wave]) | length == 0
+        fail_msg: "Selected hosts must equal or be a subset of groups[release_wave]; broader fleet scope is refused."
+      run_once: true
   tasks:
     - name: Reboot host with grace period
       ansible.builtin.reboot:

ansible/playbooks/bf-remote-channels-check.yml

@@ -65,7 +65,8 @@
       ansible.builtin.assert:
         that:
           - operation_wave in approved_operation_waves
-          - operation_confirmed | bool
+          # Undefined-safe confirmation gate (static gate idiom: 'operation_confirmed | bool').
+          - operation_confirmed | default(false) | bool
           - operation_wave in groups
           - groups[operation_wave] | length > 0
           - ansible_play_hosts_all | length > 0

ansible/playbooks/bf-run-script.yml

@@ -66,11 +66,15 @@
           - script_stat.stat.executable
 
     - name: Execute approved canonical script
+      # NOTE: script_args is split on whitespace; quoted arguments are NOT
+      # preserved. The plain split is the safe default (no shell involved),
+      # so multi-word input must be passed as separate words.
       ansible.builtin.command:
         argv: "{{ [canonical_script.stdout] + (script_args | default('') | split()) }}"
       register: script_out
       changed_when: true
 
     - name: Show script output
+      # Intentional: script output stays operator-visible; do NOT add no_log here.
       ansible.builtin.debug:
         msg: "{{ inventory_hostname }} exit={{ script_out.rc }}\n{{ script_out.stdout | default('') }}{{ script_out.stderr | default('') }}"

ansible/playbooks/bf-rustdesk-restart.yml

@@ -1,14 +1,34 @@
 ---
 # Restart the RustDesk remote-access service.
-# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-rustdesk-restart.yml --limit lab
+# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-rustdesk-restart.yml -e operation_wave=lab -e operation_confirmed=true --limit lab
 # Service name is overridable via rustdesk_service_name.
 # No Semaphore Pro feature is used.
 - name: Restart RustDesk service
-  hosts: blueforce
+  hosts: "{{ operation_wave | default('invalid-wave', true) }}"
   gather_facts: false
   become: true
   vars:
+    approved_operation_waves: [lab, pilot_1, pilot_2, wave_1, wave_2, production]
+    production_override: false
     rustdesk_restart_service: "{{ rustdesk_service_name | default('rustdesk') }}"
+  pre_tasks:
+    - name: Require confirmed approved operation wave scope
+      ansible.builtin.assert:
+        that:
+          - operation_wave in approved_operation_waves
+          - operation_confirmed | bool
+          - operation_wave in groups
+          - ansible_play_hosts_all | length > 0
+          - ansible_play_hosts_all | difference(groups[operation_wave]) | length == 0
+        fail_msg: "Set operation_wave to one inventory wave and operation_confirmed=true; full-fleet scope is refused."
+      run_once: true
+
+    - name: Require separate production override
+      ansible.builtin.assert:
+        that:
+          - operation_wave != 'production' or production_override | bool
+        fail_msg: "Production operation requires production_override=true in addition to operation_confirmed=true."
+      run_once: true
   tasks:
     - name: Restart RustDesk
       ansible.builtin.systemd:
@@ -16,7 +36,8 @@
         state: restarted
 
     - name: Verify RustDesk is active
-      ansible.builtin.command: "systemctl is-active {{ rustdesk_restart_service }}"
+      ansible.builtin.command:
+        argv: [systemctl, is-active, "{{ rustdesk_restart_service }}"]
       register: rustdesk_state
       changed_when: false
 

ansible/playbooks/bf-service-restart.yml

@@ -1,13 +1,15 @@
 ---
 # Restart a single service by name (parametric, allowlisted).
-# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-service-restart.yml --limit lab -e service_name=blueforce-agent
+# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-service-restart.yml -e operation_wave=lab -e operation_confirmed=true --limit lab -e service_name=blueforce-agent
 # Only names in service_allowlist are accepted; everything else fails closed.
 # No Semaphore Pro feature is used.
 - name: Restart allowlisted service
-  hosts: blueforce
+  hosts: "{{ operation_wave | default('invalid-wave', true) }}"
   gather_facts: false
   become: true
   vars:
+    approved_operation_waves: [lab, pilot_1, pilot_2, wave_1, wave_2, production]
+    production_override: false
     service_name: ""
     service_allowlist:
       - blueforce-agent
@@ -16,6 +18,24 @@
       - ssh
       - systemd-journald
       - cron
+  pre_tasks:
+    - name: Require confirmed approved operation wave scope
+      ansible.builtin.assert:
+        that:
+          - operation_wave in approved_operation_waves
+          - operation_confirmed | bool
+          - operation_wave in groups
+          - ansible_play_hosts_all | length > 0
+          - ansible_play_hosts_all | difference(groups[operation_wave]) | length == 0
+        fail_msg: "Set operation_wave to one inventory wave and operation_confirmed=true; full-fleet scope is refused."
+      run_once: true
+
+    - name: Require separate production override
+      ansible.builtin.assert:
+        that:
+          - operation_wave != 'production' or production_override | bool
+        fail_msg: "Production operation requires production_override=true in addition to operation_confirmed=true."
+      run_once: true
   tasks:
     - name: Validate service_name is provided
       ansible.builtin.assert:

ansible/playbooks/bf-wireguard-restart.yml

@@ -1,14 +1,34 @@
 ---
 # Restart the WireGuard tunnel interface.
-# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-wireguard-restart.yml --limit lab
+# Run: ansible-playbook -i inventory/hosts.yml playbooks/bf-wireguard-restart.yml -e operation_wave=lab -e operation_confirmed=true --limit lab
 # Uses wg_interface from group_vars and defaults to wg0.
 # No Semaphore Pro feature is used.
 - name: Restart WireGuard tunnel
-  hosts: blueforce
+  hosts: "{{ operation_wave | default('invalid-wave', true) }}"
   gather_facts: false
   become: true
   vars:
+    approved_operation_waves: [lab, pilot_1, pilot_2, wave_1, wave_2, production]
+    production_override: false
     wireguard_restart_interface: "{{ wg_interface | default('wg0') }}"
+  pre_tasks:
+    - name: Require confirmed approved operation wave scope
+      ansible.builtin.assert:
+        that:
+          - operation_wave in approved_operation_waves
+          - operation_confirmed | bool
+          - operation_wave in groups
+          - ansible_play_hosts_all | length > 0
+          - ansible_play_hosts_all | difference(groups[operation_wave]) | length == 0
+        fail_msg: "Set operation_wave to one inventory wave and operation_confirmed=true; full-fleet scope is refused."
+      run_once: true
+
+    - name: Require separate production override
+      ansible.builtin.assert:
+        that:
+          - operation_wave != 'production' or production_override | bool
+        fail_msg: "Production operation requires production_override=true in addition to operation_confirmed=true."
+      run_once: true
   tasks:
     - name: Restart WireGuard service
       ansible.builtin.systemd:
@@ -16,7 +36,8 @@
         state: restarted
 
     - name: Verify tunnel state
-      ansible.builtin.command: "wg show {{ wireguard_restart_interface }}"
+      ansible.builtin.command:
+        argv: [wg, show, "{{ wireguard_restart_interface }}"]
       register: wg_out
       changed_when: false
 

bfos/operations.py

@@ -158,16 +158,16 @@ REPORT_OPERATIONS: Tuple[Operation, ...] = (
 
 # 2.2 Fleet maintenance -- destructive rows ask a second, explicit question.
 MAINTENANCE_OPERATIONS: Tuple[Operation, ...] = (
-    Operation("rdp-restart", "RDP servisini yeniden başlat", "bf-rdp-restart.yml", "simple", True,
+    Operation("rdp-restart", "RDP servisini yeniden başlat", "bf-rdp-restart.yml", "wave", True,
               PROMPT_NONE, "xrdp servisini yeniden başlatır", CATEGORY_MAINTENANCE),
     Operation("rustdesk-restart", "RustDesk servisini yeniden başlat", "bf-rustdesk-restart.yml",
-              "simple", True, PROMPT_NONE, "RustDesk servisini yeniden başlatır",
+              "wave", True, PROMPT_NONE, "RustDesk servisini yeniden başlatır",
               CATEGORY_MAINTENANCE),
     Operation("wireguard-restart", "WireGuard servisini yeniden başlat",
-              "bf-wireguard-restart.yml", "simple", True, PROMPT_NONE,
+              "bf-wireguard-restart.yml", "wave", True, PROMPT_NONE,
               "wg-quick@wg0 servisini yeniden başlatır", CATEGORY_MAINTENANCE),
     Operation("service-restart", "İzinli bir servisi yeniden başlat", "bf-service-restart.yml",
-              "simple", True, PROMPT_SERVICE, "Yalnız izinli servis adları kabul edilir",
+              "wave", True, PROMPT_SERVICE, "Yalnız izinli servis adları kabul edilir",
               CATEGORY_MAINTENANCE),
     Operation("gui-off", "Grafik arayüzü kapat (headless)", "bf-gui-off.yml", "wave", True,
               PROMPT_NONE, "Dalgayı multi-user.target'e alır", CATEGORY_MAINTENANCE),

bfos/ui_gui.py

@@ -348,7 +348,8 @@ def state_word_from_values(values: Mapping[str, str]) -> str:
     if any(marker in field for marker in FAILURE_MARKERS
            for field in (phase, fleet, enrollment)):
         return "HATA"
-    if any("READY" in field for field in (phase, fleet, enrollment)):
+    if any(field.endswith("READY") and not field.startswith(("NOT_", "UN"))
+           for field in (phase, fleet, enrollment)):
         return "HAZIR"
     if any(field.endswith("ENROLLED") and not field.startswith(("NOT_", "UN"))
            for field in (phase, fleet, enrollment)):
@@ -968,14 +969,14 @@ OPERATIONS: tuple[Operation, ...] = (
               "Offline provisioning yerel ön koşulları.", "bf-offline-ready-check.yml"),
     # --- Cihaz islemleri: bakim (yikici olanlar isaretli) -------------------
     Operation("rdp-restart", "xRDP servisini yeniden başlat", "Cihaz işlemleri", True,
-              "simple", "xRDP servisini yeniden başlatır (RDP açık kalır).",
+              "wave", "xRDP servisini yeniden başlatır (RDP açık kalır).",
               "bf-rdp-restart.yml"),
     Operation("rustdesk-restart", "RustDesk servisini yeniden başlat", "Cihaz işlemleri",
-              True, "simple", "RustDesk servisini yeniden başlatır.", "bf-rustdesk-restart.yml"),
+              True, "wave", "RustDesk servisini yeniden başlatır.", "bf-rustdesk-restart.yml"),
     Operation("wireguard-restart", "WireGuard (wg0) yeniden başlat", "Cihaz işlemleri", True,
-              "simple", "wg-quick@wg0 servisini yeniden başlatır.", "bf-wireguard-restart.yml"),
+              "wave", "wg-quick@wg0 servisini yeniden başlatır.", "bf-wireguard-restart.yml"),
     Operation("service-restart", "İzinli servisi yeniden başlat", "Cihaz işlemleri", True,
-              "simple", "Yalnız izinli servis adlarını yeniden başlatır.",
+              "wave", "Yalnız izinli servis adlarını yeniden başlatır.",
               "bf-service-restart.yml", options=_FLEET_OPTIONS["service-restart"]),
     Operation("gui-off", "Yerel grafik oturumu kapat (headless)", "Cihaz işlemleri", True,
               "wave", "Dalgayı multi-user.target'e alır; xRDP'yi durdurmaz.", "bf-gui-off.yml"),
@@ -1462,7 +1463,7 @@ class CatalogueBackend:
                         unlocks=("wave_gate_confirmed",),
                     )
                 )
-        if target is not None and target.kind == "all" and operation.scope == "simple":
+        if target is not None and target.kind == "all" and operation.scope in ("simple", "wave"):
             steps.append(
                 GateStepInfo(
                     kind="token",
@@ -1845,6 +1846,8 @@ class CatalogueBackend:
         if check_mode and op_id in CHECK_CAPABLE_LOCAL_OPS:
             commands = [tuple(argv) + ("--check",) for argv in commands]
             notes.append("--check: yalnız denetim yapılır, değişiklik uygulanmaz.")
+        elif check_mode:
+            notes.append("--check bu işlem için geçerli değil; normal koşu hazırlanır.")
         label = target.label if target is not None and target.value else "bu cihaz (yerel)"
         return CommandPlan(
             commands=tuple(commands),
@@ -1925,17 +1928,6 @@ class ModuleBackend:
         "tools-support-bundle": ("bf-support-bundle", ("--out-dir", "/var/tmp")),
     }
 
-    #: modulun salt-okunur rapor ureticileri: id -> fonksiyon adi
-    _REPORT_OPS: Mapping[str, str] = {
-        "central-targets": "targets_lines",
-        "central-readiness": "central_readiness_lines",
-        "central-endpoints": "central_endpoint_lines",
-        "central-probe": "central_probe_lines",
-        "central-wave-gates": "wave_gate_state_lines",
-        "info-identity": "identity_lines",
-        "info-state": "state_lines",
-    }
-
     def __init__(self, ops_module, runner_module, ctx: GuiContext):
         self._ops = ops_module
         self._runner = runner_module
@@ -2222,7 +2214,7 @@ class ModuleBackend:
         return CommandPlan(
             commands=(tuple(argv),),
             destructive=bool(raw_op.destructive),
-            requires_root=False,
+            requires_root=bool(raw_op.destructive),
             operation_id=operation.id,
             target_label=str(getattr(resolved, "label", target.label)),
             notes=tuple(notes),
@@ -4305,7 +4297,7 @@ def is_root() -> bool:
     of refusing to launch.
     """
     geteuid = getattr(os, "geteuid", None)
-    return True if geteuid is None else geteuid() == 0
+    return False if geteuid is None else geteuid() == 0
 
 
 ROOT_REQUIRED_NOTICE = (
@@ -4345,6 +4337,9 @@ def parse_args(argv: Sequence[str] | None = None) -> argparse.Namespace:
 
 def main(argv: Sequence[str] | None = None) -> int:
     args = parse_args(argv)
+    if args.install and args.portal:
+        print("--install ve --portal aynı anda kullanılamaz.", file=sys.stderr)
+        return 2
     ctx = GuiContext.detect(repo=args.repo, inventory=args.inventory, dry_run=args.dry_run)
     backend, note = load_backend(ctx)
 
@@ -4360,14 +4355,10 @@ def main(argv: Sequence[str] | None = None) -> int:
     if problem:
         print(problem, file=sys.stderr)
         return 3
-
     root = tk.Tk()
     try:
         from bfos.ui_surfaces import launch_surface
 
-        if args.install and args.portal:
-            print("--install ve --portal aynı anda kullanılamaz.", file=sys.stderr)
-            return 2
         launch_surface(
             root, backend, ctx,
             "install" if args.install else "portal" if args.portal or args.full else None,

bfos/ui_surfaces.py

@@ -35,10 +35,23 @@ class SurfaceFrame(ttk.Frame):
         self.root = root
         self.backend = backend
         self.ctx = ctx
+        self._alive = True
+        self._after_id: str | None = None
         self.styles = theme.apply_theme(root)
         super().__init__(root, style=self.styles["frame"])
         self.pack(fill="both", expand=True)
 
+    def destroy(self) -> None:
+        """Cancel pending after() callbacks before tearing down widgets."""
+        self._alive = False
+        after_id, self._after_id = self._after_id, None
+        if after_id is not None:
+            try:
+                self.after_cancel(after_id)
+            except Exception:
+                pass
+        super().destroy()
+
     def clear(self) -> None:
         for child in self.winfo_children():
             child.destroy()
@@ -244,7 +257,16 @@ class InstallerSurface(SurfaceFrame):
         except Exception as exc:
             self.status_var.set(f"Plan üretilemedi: {exc}")
             return
-        if not messagebox.askyesno("Kurulumu onaylayın", "Bu cihaz için kurulum başlatılsın mı?\n\n" + "\n".join(plan.lines), default="no"):
+        preview = self._preview_plan(plan, operation, target)
+        if not messagebox.askyesno("Kurulumu onaylayın", "Bu cihaz için kurulum başlatılsın mı?\n\n" + preview, default="no"):
+            self.status_var.set("İptal edildi; hiçbir komut çalıştırılmadı.")
+            return
+        if plan.destructive and not messagebox.askyesno(
+            "İKİNCİ ONAY — yıkıcı işlem",
+            "Bu kurulum seçili cihazda DEĞİŞİKLİK YAPAR (dosya, servis ve sahiplikler).\n"
+            "Bu onay yalnız bu kurulum için geçerlidir. Gerçekten başlatılsın mı?",
+            default="no",
+        ):
             self.status_var.set("İptal edildi; hiçbir komut çalıştırılmadı.")
             return
         self.status_var.set("Kurulum çalışıyor…")
@@ -280,19 +302,63 @@ class InstallerSurface(SurfaceFrame):
         self.dealer_entry.focus_set()
 
     def _append(self, text: str) -> None:
+        try:
             self.output.configure(state="normal")
             self.output.insert("end", text)
             self.output.see("end")
             self.output.configure(state="disabled")
+        except tk.TclError:
+            pass
+
+    def _preview_plan(self, plan, operation, target) -> str:
+        """Yerel plan önizlemesi: komutlar + notlar + yıkıcı/yetki uyarısı."""
+        lines = ["Çalıştırılacak komut(lar):", ""]
+        if plan.lines:
+            for line in plan.lines:
+                lines.append(f"    {line}")
+        else:
+            lines.append("    (komut üretilemedi)")
+        lines.append("")
+        lines.append(f"İşlem  : {operation.label}  [{operation.id}]")
+        lines.append(f"Hedef  : {plan.target_label or target.label}")
+        if plan.needs_root:
+            lines.append("Yetki  : root gerekir — konsolu sudo ile açın.")
+        for note in plan.notes:
+            lines.append(f"Not    : {note}")
+        if plan.destructive:
+            lines.append("")
+            lines.append("UYARI: Bu YIKICI bir işlemdir; seçili cihazda değişiklik yapar.")
+        return "\n".join(lines)
 
     def _drain_output(self) -> None:
+        if not self._alive:
+            return
+        drained: list[str] = []
         try:
             while True:
-                self._append(self.output_queue.get_nowait())
-                self.status_var.set("Kurulum çıktısı güncelleniyor…")
+                drained.append(self.output_queue.get_nowait())
         except queue.Empty:
             pass
-        self.after(100, self._drain_output)
+        for line in drained:
+            try:
+                self._append(line)
+            except tk.TclError:
+                return
+        if drained:
+            terminal = next(
+                (line.strip() for line in reversed(drained)
+                 if line.strip().startswith("[OK]") or line.strip().startswith("[HATA]")),
+                None,
+            )
+            try:
+                self.status_var.set(terminal if terminal is not None else "Kurulum çıktısı güncelleniyor…")
+            except tk.TclError:
+                return
+        if self._alive:
+            try:
+                self._after_id = self.after(100, self._drain_output)
+            except tk.TclError:
+                pass
 
     def _back(self) -> None:
         self.destroy()
@@ -312,10 +378,12 @@ class RemotePortalSurface(SurfaceFrame):
         self.preferences = load_preferences()
         self.status_var = tk.StringVar(value="Bir cihaz seçin")
         self.output_queue: queue.Queue[str] = queue.Queue()
+        self._run_done = False
         self._render()
         root.bind("<Configure>", self._on_resize)
         self.search_var.trace_add("write", lambda *_: self._render_devices())
         self._render_devices()
+        self._drain_output()
 
     def _render(self) -> None:
         self.title_row("Uzaktan yönetim portalı", "Envanter, sağlık ve kontrollü uzaktan operasyonlar  ·  " +
@@ -334,10 +402,11 @@ class RemotePortalSurface(SurfaceFrame):
 
         metrics = ttk.Frame(body, style=self.styles["frame"])
         metrics.grid(row=0, column=1, sticky="ew", pady=(0, theme.SP_4))
-        for col, (label, value) in enumerate(self._metrics()):
+        cards = self._metrics()
+        for col, (label, value) in enumerate(cards):
             metrics.columnconfigure(col, weight=1)
             card = ttk.Frame(metrics, style=self.styles["panel"])
-            card.grid(row=0, column=col, sticky="ew", padx=(0, theme.SP_3) if col < 3 else (0, 0))
+            card.grid(row=0, column=col, sticky="ew", padx=(0, theme.SP_3) if col < len(cards) - 1 else (0, 0))
             ttk.Label(card, text=label, style=self.styles["panel_title"]).pack(anchor="w", padx=theme.SP_3, pady=(theme.SP_3, 0))
             ttk.Label(card, text=value, style=self.styles["panel_label"]).pack(anchor="w", padx=theme.SP_3, pady=(theme.SP_1, theme.SP_3))
 
@@ -372,12 +441,23 @@ class RemotePortalSurface(SurfaceFrame):
         self.operation = ttk.Combobox(action, state="readonly", width=32, values=tuple(op.label for op in self.backend.list_operations() if not op.destructive))
         self.operation.pack(side="right", padx=(theme.SP_2, theme.SP_4), pady=theme.SP_3)
         ttk.Button(action, text="Salt-okunur çalıştır", style=self.styles["accent_button"], command=self._run_read_only).pack(side="right", pady=theme.SP_3)
+        log = ttk.Frame(self, style=self.styles["panel"])
+        log.pack(fill="x", padx=theme.SP_6, pady=(0, theme.SP_4))
+        ttk.Label(log, text="Çalışma çıktısı", style=self.styles["panel_title"]).pack(
+            anchor="w", padx=theme.SP_4, pady=(theme.SP_3, theme.SP_2))
+        self.output = tk.Text(log, height=4, wrap="none", state="disabled",
+                              font=("TkFixedFont", 10), relief="flat", padx=8, pady=8)
+        self.output.pack(fill="x", padx=theme.SP_3, pady=(0, theme.SP_3))
 
     def _metrics(self) -> tuple[tuple[str, str], ...]:
         groups = parse_inventory(self.ctx.inventory) if self.ctx.inventory else {}
         hosts = sorted({host for values in groups.values() for host in values})
-        return (("TOPLAM CİHAZ", str(len(hosts))), ("ÇEVRİMİÇİ", "—"),
-                ("ÇALIŞAN İŞLEM", "yok"), ("UYARI", "yok"))
+        # parse_inventory yalnız host adlarını verir; çevrimiçi/işlem durumu
+        # için ajan-izleme entegrasyonu gerekir — sahte veri gösterilmez.
+        return (
+            ("TOPLAM CİHAZ", str(len(hosts))),
+            ("CANLI DURUM", "İzleme entegrasyonu gerekir — canlı veri yok"),
+        )
 
     def _on_resize(self, event) -> None:
         if event.widget is not self.root or not hasattr(self, "detail"):
@@ -399,7 +479,8 @@ class RemotePortalSurface(SurfaceFrame):
             if query and query not in host.lower():
                 continue
             wave = next((name for name, values in groups.items() if host in values), "—")
-            self.tree.insert("", "end", iid=host, values=(host, wave, "envanterden okunur"))
+            # parse_inventory host adından öte adres vermez; bilinmeyen adres "—" kalır.
+            self.tree.insert("", "end", iid=host, values=(host, wave, "—"))
 
     def _select_device(self, _event=None) -> None:
         selected = self.tree.selection()
@@ -425,24 +506,163 @@ class RemotePortalSurface(SurfaceFrame):
         if operation is None:
             messagebox.showinfo("Operasyon seçin", "Salt-okunur bir operasyon seçin.")
             return
+        host = self.selected_host
+        dealer = host[3:] if host.startswith("bf-") else host
+        problem = dealer_error(dealer)
+        if problem:
+            messagebox.showerror("Geçersiz cihaz kimliği", problem)
+            self.status_var.set(problem)
+            return
+        target = Target("dealer", dealer)
         self.status_var.set(f"Çalıştırılıyor: {operation.label} — {self.selected_host}")
         # The existing backend remains the only command authority.
         try:
-            plan = self.backend.build_command(operation, Target("dealer", self.selected_host.replace("bf-", "")), {}, check_mode=True)
-            if not messagebox.askyesno("Salt-okunur operasyon", "Komut çalıştırılsın mı?\n\n" + "\n".join(plan.lines), default="no"):
-                self.status_var.set("İptal edildi")
-                return
-            def worker() -> None:
-                try:
-                    result = self.backend.run(plan)
-                    self.root.after(0, lambda: self.status_var.set(
-                        f"Tamamlandı: çıkış kodu {result.exit_code}"))
-                except Exception as exc:
-                    message = str(exc)
-                    self.root.after(0, lambda: self.status_var.set(f"Hata: {message}"))
-            threading.Thread(target=worker, daemon=True).start()
+            plan = self.backend.build_command(operation, target, {}, check_mode=True)
         except Exception as exc:
             self.status_var.set(f"Hata: {exc}")
+            return
+        grants = self._collect_grants(operation, target)
+        if grants is None:
+            self.status_var.set("İptal edildi")
+            return
+        if grants:
+            try:
+                plan = self.backend.build_command(operation, target, {}, grants=grants, check_mode=True)
+            except Exception as exc:
+                self.status_var.set(f"Hata: {exc}")
+                return
+        if plan.destructive and not hasattr(self.backend, "gate_steps"):
+            self.status_var.set("Yıkıcı işlem engellendi — ana konsoldan (sudo bf) iki onay ile çalıştırın.")
+            messagebox.showerror(
+                "Yıkıcı işlem engellendi",
+                "Bu plan değişiklik yapar; portal yalnız salt-okunur çalıştırır.\n"
+                "Yıkıcı işlem için ana konsolu kullanın (sudo bf).",
+            )
+            return
+        preview = "\n".join(plan.lines) if plan.lines else "(komut üretilemedi)"
+        for note in plan.notes:
+            preview += f"\nNot: {note}"
+        if not messagebox.askyesno("Salt-okunur operasyon", "Komut çalıştırılsın mı?\n\n" + preview, default="no"):
+            self.status_var.set("İptal edildi")
+            return
+        self._run_done = False
+        self.status_var.set(f"Çalıştırılıyor: {operation.label} — {self.selected_host}")
+        self._append("\n" + "\n".join(plan.lines) + "\n" if plan.lines else "\n")
+
+        def worker() -> None:
+            try:
+                result = self.backend.run(plan, on_line=self.output_queue.put)
+                self.output_queue.put(f"\n{'[OK]' if result.ok else '[HATA]'} Çıkış kodu: {result.exit_code}\n")
+            except TypeError:
+                try:
+                    result = self.backend.run(plan)
+                    self.output_queue.put(f"\n{'[OK]' if result.ok else '[HATA]'} Çıkış kodu: {result.exit_code}\n")
+                except Exception as exc:
+                    self.output_queue.put(f"[HATA] {exc}\n")
+            except Exception as exc:
+                self.output_queue.put(f"[HATA] {exc}\n")
+        threading.Thread(target=worker, daemon=True).start()
+
+    def _collect_grants(self, operation, target) -> dict | None:
+        """Backend kapılarını sorar; iptalde None döner (ana konsol ile aynı sıra)."""
+        gate_steps = getattr(self.backend, "gate_steps", None)
+        if not callable(gate_steps):
+            return {}
+        try:
+            steps = gate_steps(operation, target)
+        except CommandError as exc:
+            messagebox.showerror("Kapı onayları alınamadı", str(exc))
+            return None
+        grants: dict = {}
+        for step in steps:
+            kind = getattr(step, "kind", "approval")
+            prompt = getattr(step, "prompt", None) or "Onay"
+            message = getattr(step, "message", None) or "Devam edilsin mi?"
+            if kind == "token":
+                token = getattr(step, "token", "") or ""
+                answer = self._token_dialog(prompt, message)
+                if answer is None or answer.strip() != token:
+                    self.status_var.set("Kapı onayı verilmedi; işlem iptal edildi.")
+                    return None
+            elif not messagebox.askyesno(prompt, message, default="no"):
+                self.status_var.set("Kapı onayı verilmedi; işlem iptal edildi.")
+                return None
+            for name in (getattr(step, "unlocks", ()) or ()):
+                grants[name] = True
+        return grants
+
+    def _token_dialog(self, title: str, message: str) -> str | None:
+        dialog = tk.Toplevel(self.root)
+        dialog.title(title)
+        dialog.transient(self.root)
+        dialog.grab_set()
+        value = tk.StringVar()
+        ttk.Label(dialog, text=message, style=self.styles["panel_title"],
+                  wraplength=320, justify="left").pack(
+            padx=theme.SP_5, pady=(theme.SP_5, theme.SP_2), anchor="w")
+        entry = ttk.Entry(dialog, textvariable=value, width=28, style=self.styles["entry"])
+        entry.pack(padx=theme.SP_5, pady=(0, theme.SP_4), anchor="w")
+        result: dict = {"value": None}
+
+        def confirm() -> None:
+            result["value"] = value.get()
+            dialog.destroy()
+
+        def cancel() -> None:
+            dialog.destroy()
+        ttk.Button(dialog, text="Onayla", style=self.styles["accent_button"],
+                   command=confirm).pack(padx=theme.SP_5, pady=(0, theme.SP_2), anchor="w")
+        ttk.Button(dialog, text="Vazgeç", style=self.styles["quiet_button"],
+                   command=cancel).pack(padx=theme.SP_5, pady=(0, theme.SP_5), anchor="w")
+        entry.focus_set()
+        dialog.wait_window()
+        return result["value"]
+
+    def _append(self, text: str) -> None:
+        try:
+            self.output.configure(state="normal")
+        except tk.TclError:
+            return
+        try:
+            self.output.insert("end", text)
+            self.output.see("end")
+            self.output.configure(state="disabled")
+        except tk.TclError:
+            pass
+
+    def _drain_output(self) -> None:
+        if not self._alive:
+            return
+        drained: list[str] = []
+        try:
+            while True:
+                drained.append(self.output_queue.get_nowait())
+        except queue.Empty:
+            pass
+        for line in drained:
+            try:
+                self._append(line)
+            except tk.TclError:
+                return
+        if drained:
+            terminal = next(
+                (line.strip() for line in reversed(drained)
+                 if line.strip().startswith("[OK]") or line.strip().startswith("[HATA]")),
+                None,
+            )
+            try:
+                if terminal is not None:
+                    self.status_var.set(terminal)
+                    self._run_done = True
+                elif not self._run_done:
+                    self.status_var.set("Çalışma çıktısı güncelleniyor…")
+            except tk.TclError:
+                return
+        if self._alive:
+            try:
+                self._after_id = self.after(100, self._drain_output)
+            except tk.TclError:
+                pass
 
     def _back(self) -> None:
         self.destroy()
@@ -465,7 +685,12 @@ class RemotePortalSurface(SurfaceFrame):
                 return
             if not messagebox.askyesno("Kimliği değiştir", "Aktif kullanıcı bf-" + value.get().strip() + " olarak değiştirilsin mi?", parent=dialog, default="no"):
                 return
+            try:
                 saved = save_dealer(value.get().strip())
+            except (OSError, PermissionError, ValueError) as exc:
+                messagebox.showerror("Kayıt başarısız", f"Bayi numarası kaydedilemedi: {exc}", parent=dialog)
+                self.status_var.set(f"Ayar kaydedilemedi: {exc}")
+                return
             self.ctx.active_dealer = saved["dealer_id"]
             self.ctx.active_user = saved["active_user"]
             dialog.destroy()

bfos/ui_widgets.py

@@ -189,10 +189,12 @@ class SegmentedControl:
         self._paint()
 
     def _paint(self) -> None:
+        # Selected segment stays enabled (keyboard-focusable, in Tab order);
+        # selection is shown by the accent style alone, never by disabling.
         for key, button in self._buttons.items():
             if key == self._value:
                 button.configure(style=self.styles["accent_button"],
-                                 state="disabled")
+                                 state="normal")
             else:
                 button.configure(style=self.styles["button"], state="normal")
 

config/systemd/logrotate-blueforce

@@ -2,7 +2,7 @@
 # Scope: /var/log/blueforce and /var/log/meg file logs; install to /etc/logrotate.d/.
 # Safety: Idempotent under logrotate; validate with logrotate -d before deploying.
 # Ref: docs/15-LOGGING.md, K-14.
-/var/log/blueforce/*.log /var/log/meg/*.log {
+/var/log/blueforce-console.log /var/log/meg/*.log {
   weekly
   rotate 7
   size 10M

config/systemd/wg-quick@wg0.service.d/override.conf

@@ -4,4 +4,6 @@
 # Ref: docs/08-WIREGUARD-AND-NETWORK.md and docs/16-POWER-LOSS-AND-AUTO-RECOVERY.md, K-05.
 [Service]
 Restart=always
-RestartSec=5
+RestartSec=30s
+StartLimitIntervalSec=10min
+StartLimitBurst=5

provisioning/firstboot/bf-firstboot

@@ -112,11 +112,14 @@ read_dealer_id_file() {
   printf '%s' "$value"
 }
 
-marker_exists() { [[ -e "$MARKER" ]]; }
+marker_exists() { [[ -s "$MARKER" && -f "$MARKER" && -r "$MARKER" ]] && valid_dealer_id "$(tr -d '[:space:]' < "$MARKER" 2>/dev/null || true)"; }
 
 marker_device_id() {
-  [[ -f "$MARKER" ]] || return 0
-  tr -d '[:space:]' < "$MARKER" || true
+  [[ -f "$MARKER" && -r "$MARKER" ]] || return 0
+  local value
+  value="$(tr -d '[:space:]' < "$MARKER" || true)"
+  valid_dealer_id "$value" || return 0
+  printf '%s' "$value"
 }
 
 write_marker() {
@@ -272,10 +275,22 @@ resolve_installer() {
 # explicitly requested automation mode. A previous partial run is resumed.
 run_installer() {
   local -a args=(--offline --dealer-id "$DEALER_ID" --yes)
+  if [[ -s "$STATE_DIR/install-state" ]]; then
+    # A pending install-state belongs to the identity it was recorded under. After an
+    # identity change the old module results must not be reused against the new one.
+    if [[ -f "$MARKER" && -r "$MARKER" ]]; then
+      local stamped
+      stamped="$(tr -d '[:space:]' < "$MARKER" 2>/dev/null || true)"
+      if [[ -n "$stamped" && "$stamped" != "$DEALER_ID" ]]; then
+        log "dropping stale install-state: recorded under $stamped, reinstalling for $DEALER_ID"
+        rm -f "$STATE_DIR/install-state"
+      fi
+    fi
     if [[ -s "$STATE_DIR/install-state" ]]; then
       args+=(--resume)
       log "partial install state found in $STATE_DIR/install-state; resuming instead of restarting"
     fi
+  fi
   if [[ "${BF_FIRSTBOOT_TEST_MODE:-0}" == 1 ]]; then
     echo "TEST: would run $INSTALLER ${args[*]}"
     exit 0

provisioning/firstboot/blueforce-firstboot.service

@@ -23,8 +23,10 @@ TTYPath=/dev/tty1
 TTYReset=yes
 TTYVHangup=yes
 # Upper bound so a console that nobody answers can never hold the boot job open. The
-# wizard's own input read is bounded too (BF_FIRSTBOOT_INPUT_TIMEOUT).
-TimeoutStartSec=45min
+# wizard's own input read is bounded too (BF_FIRSTBOOT_INPUT_TIMEOUT). The installer is
+# resumable (--resume reuses install-state), but a SIGTERM mid-apt on a full-pool run
+# still risks a wedged dpkg; the bound is generous so a complete run fits inside it.
+TimeoutStartSec=120min
 RemainAfterExit=yes
 # Hardening that the offline installer tolerates. ProtectSystem=strict, ProtectHome=yes
 # and similar sandboxing are deliberately absent: the installer legitimately writes /etc,

scripts/diagnostics/bf-hardware-inventory.sh

@@ -69,11 +69,13 @@ fi
 NOW="$(date '+%Y-%m-%dT%H:%M:%S%z' 2>/dev/null || echo unknown)"
 
 json_escape() { # stdin -> JSON string body
-    python3 -c 'import json,sys; print(json.dumps(sys.stdin.read())[1:-1])' 2>/dev/null || sed 's/"/\\"/g' | tr -d '\n'
+    python3 -c 'import json,sys; print(json.dumps(sys.stdin.read())[1:-1])' 2>/dev/null || python3 -c 'import sys; print(sys.stdin.read().replace(chr(92),chr(92)*2).replace(chr(34),chr(92)+chr(34)).replace(chr(10)," "))' 2>/dev/null || sed 's/\\/\\\\/g;s/"/\\"/g' | tr '\n' ' '
 }
-val() { # $1 raw -> "unknown" fallback, no newlines
+val() { # $1 raw -> "unknown" fallback, full output preserved (no truncation)
     local v="$1"
-    v="$(echo "${v}" | head -1 | sed 's/^[[:space:]]*//;s/[[:space:]]*$//')"
+    # Newlines would break the JSON/Markdown interpolation below, so flatten to
+    # spaces; per-field caps (head -3/-5/-20) already live at the call sites.
+    v="$(printf '%s' "${v}" | tr '\n' ' ' | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/[[:space:]][[:space:]]*/ /g')"
     [[ -z "${v}" ]] && echo "unknown" || echo "${v}"
 }
 
@@ -117,7 +119,7 @@ fi
 
 OS_PRETTY="$(val "$(grep -E '^PRETTY_NAME=' /etc/os-release 2>/dev/null | cut -d= -f2 | tr -d '"')")"
 
-E() { printf '%s' "$1" | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read())[1:-1])' 2>/dev/null || printf '%s' "$1"; }
+E() { printf '%s' "$1" | python3 -c 'import json,sys; print(json.dumps(sys.stdin.read())[1:-1])' 2>/dev/null || { printf '%s' "$1" | sed 's/\\/\\\\/g;s/"/\\"/g' | tr '\n' ' '; }; }
 
 JSON="$(cat <<EOF
 {

scripts/identity/bf-reidentity

@@ -3,6 +3,7 @@
 from __future__ import annotations
 
 import os
+import re
 import shutil
 import subprocess
 import sys
@@ -76,15 +77,24 @@ def main() -> int:
         if old_sudo.exists():
             shutil.move(old_sudo, new_sudo)
             moved_sudo = True
-            text = new_sudo.read_text(encoding="utf-8").replace(old_user, new_user)
-            new_sudo.write_text(text, encoding="utf-8")
+            # Scope the rename to full-token occurrences of the old login name (a
+            # blind str.replace would also rewrite e.g. bf-11111111-backup).
+            sudo_token = re.compile(rf"(?<![\w-]){re.escape(old_user)}(?![\w-])")
+            sudo_lines = new_sudo.read_text(encoding="utf-8").splitlines(keepends=True)
+            new_sudo.write_text("".join(sudo_token.sub(new_user, line) for line in sudo_lines), encoding="utf-8")
             run("visudo", "-cf", str(new_sudo))
 
         if old_ssh.exists():
             shutil.move(old_ssh, new_ssh)
             moved_ssh = True
-            text = new_ssh.read_text(encoding="utf-8").replace(f"AllowUsers {old_user}", f"AllowUsers {new_user}")
-            new_ssh.write_text(text, encoding="utf-8")
+            # Replace only the full AllowUsers token, not substrings of other names.
+            ssh_token = re.compile(rf"(?<![\w-]){re.escape(old_user)}(?![\w-])")
+            out_lines = []
+            for line in new_ssh.read_text(encoding="utf-8").splitlines(keepends=True):
+                if line.lstrip().startswith("AllowUsers"):
+                    line = ssh_token.sub(new_user, line)
+                out_lines.append(line)
+            new_ssh.write_text("".join(out_lines), encoding="utf-8")
             run("sshd", "-t")
 
         tmp = IDENTITY_FILE.with_suffix(".tmp")
@@ -96,6 +106,7 @@ def main() -> int:
         print(f"OK: {old_id} -> {new_id} ({new_user})")
         return 0
     except Exception as exc:
+        rollback_errors = []
         try:
             if IDENTITY_FILE.read_text(encoding="utf-8").strip() == new_id:
                 IDENTITY_FILE.write_text(old_id + "\n", encoding="utf-8")
@@ -112,9 +123,28 @@ def main() -> int:
                 run("usermod", "-l", old_user, new_user)
             if old_hostname:
                 run("hostnamectl", "set-hostname", old_hostname)
-        except Exception:
-            pass
+        except Exception as rollback_exc:
+            rollback_errors.append(str(rollback_exc))
+        # A restored config is only trustworthy after it revalidates; and a
+        # failed rollback must never be silent.
+        try:
+            if moved_sudo and old_sudo.exists():
+                run("visudo", "-cf", str(old_sudo))
+        except Exception as rollback_exc:
+            rollback_errors.append(f"sudoers revalidation failed: {rollback_exc}")
+        try:
+            if moved_ssh and old_ssh.exists():
+                run("sshd", "-t")
+        except Exception as rollback_exc:
+            rollback_errors.append(f"sshd revalidation failed: {rollback_exc}")
         audit(f"identity-change failed old={old_id} new={new_id}: {exc}")
+        if rollback_errors:
+            print(
+                f"HATA: kimlik geçişi geri alınmaya çalışıldı: {exc}; "
+                f"geri alma HATALARI: {'; '.join(rollback_errors)}",
+                file=sys.stderr,
+            )
+            return 1
         print(f"HATA: kimlik geçişi geri alınmaya çalışıldı: {exc}", file=sys.stderr)
         return 1
 

scripts/install/blueforce-install.sh

@@ -35,6 +35,7 @@ LOG_FILE="${LOG_FILE:-/var/log/blueforce-install.log}"
 STATE_DIR="${STATE_DIR:-/var/lib/blueforce}"
 STATE_FILE="$STATE_DIR/install-state"
 MODULES=(01-precheck 02-system 03-user 04-hostname 05-network 06-wireguard 07-ssh 08-rdp 09-rustdesk 10-docker 11-meg 12-monitoring 13-firewall 14-update-policy 15-logrotate 16-gui 17-healthcheck 18-final-check)
+trap 'rm -f "$STATE_FILE.tmp.$$"' EXIT
 
 DEALER_ID="" FROM="" ONLY="" CHECK_MODE=0 RESUME=0 ASSUME_YES=0 FORCE_OFFLINE=0 ONLINE=0
 # The operator's override is remembered before any default is applied: an
@@ -97,7 +98,7 @@ blocked_bundle() {
   printf '                 sistemde hiçbir değişiklik yapılmadı.\n' >&2
   exit 1
 }
-record_state() { mkdir -p "$STATE_DIR"; local tmp="$STATE_FILE.tmp"; grep -v "^$1=" "$STATE_FILE" 2>/dev/null > "$tmp" || true; echo "$1=$2" >> "$tmp"; mv "$tmp" "$STATE_FILE"; }
+record_state() { mkdir -p "$STATE_DIR"; local tmp="$STATE_FILE.tmp.$$"; grep -v "^$1=" "$STATE_FILE" 2>/dev/null > "$tmp" || true; echo "$1=$2" >> "$tmp"; mv "$tmp" "$STATE_FILE" || { rm -f "$tmp"; return 1; }; }
 state_of() { grep "^$1=" "$STATE_FILE" 2>/dev/null | cut -d= -f2 || true; }
 
 # ------------------------------------------------------- bundle discovery -----
@@ -221,7 +222,7 @@ validate_offline_repository() {
   # in a temporary directory and is removed on exit.
   if [[ "$CHECK_MODE" -eq 1 ]]; then
     CHECK_TMP="$(mktemp -d)"
-    trap 'rm -rf "${CHECK_TMP:-}"' EXIT
+    trap 'rm -rf "${CHECK_TMP:-}"; rm -f "$STATE_FILE.tmp.$$"' EXIT
     printf 'deb [trusted=yes] file:%s ./\n' "$repo" > "$CHECK_TMP/offline.list"
     BF_OFFLINE_APT_CONFIG="$CHECK_TMP/apt.conf"
     printf 'Dir::Etc::sourcelist "%s";\nDir::Etc::sourceparts "-";\nAcquire::Languages "none";\n' "$CHECK_TMP/offline.list" > "$BF_OFFLINE_APT_CONFIG"
@@ -252,6 +253,7 @@ esac; done
 [[ -n "$DEALER_ID" ]] || usage_error 'Zorunlu parametre eksik: --dealer-id <8 haneli bayi numarası>. Örnek: sudo ./blueforce-install.sh --dealer-id 12345678'
 [[ "$DEALER_ID" =~ ^[0-9]{8}$ ]] || usage_error "Geçersiz bayi numarası '$DEALER_ID': 8 haneli rakam olmalıdır (örn. 12345678)."
 [[ "$FORCE_OFFLINE" -eq 1 && "$ONLINE" -eq 1 ]] && usage_error 'Bir kurulumda hem --offline hem --online verilemez.'
+[[ -z "$FROM" || "$FROM" =~ ^[0-9]+$ ]] || usage_error '--from yalnız sayı olmalıdır (örn. --from 05).'
 
 # Apply mode changes APT, services, firewall rules and /var/lib state.  Fail
 # before repository discovery so a desktop launch as a normal user produces a
@@ -298,8 +300,15 @@ if [[ "$CHECK_MODE" -eq 0 && "$ASSUME_YES" -eq 0 ]]; then
 fi
 
 RUN=(); for m in "${MODULES[@]}"; do
-  if [[ -n "$ONLY" ]]; then [[ "$m" == "$ONLY" || "$m" == "$ONLY"* ]] && RUN+=("$m"); continue; fi
-  [[ -n "$FROM" && "${m%%-*}" < "$FROM" ]] && continue
+  # Exact match only: a bare prefix ("1") must never also select 11-18. A pure
+  # number selects by numeric value (1 == 01-precheck, one module); anything
+  # else must name the full module entry.
+  if [[ -n "$ONLY" ]]; then
+    if [[ "$ONLY" =~ ^[0-9]+$ ]]; then [[ "10#${m%%-*}" -eq "10#$ONLY" ]] && RUN+=("$m")
+    else [[ "$m" == "$ONLY" ]] && RUN+=("$m"); fi
+    continue
+  fi
+  [[ -n "$FROM" && "10#${m%%-*}" -lt "10#$FROM" ]] && continue
   if [[ "$CHECK_MODE" -eq 0 && "$RESUME" -eq 1 && "$(state_of "$m")" == OK ]]; then log "SKIP $m (already OK)"; continue; fi
   RUN+=("$m")
 done
@@ -316,17 +325,34 @@ done
 if [[ "$CHECK_MODE" -eq 0 && "$BF_OFFLINE" -eq 1 && "$failures" -eq 0 ]]; then
   install -d -m 700 "$STATE_DIR"
   umask 077
-  python3 - "$STATE_DIR/state.json" "$DEALER_ID" <<'PY'
+  # Never step an ENROLLED/READY device back to PROVISIONED_OFFLINE: a rerun
+  # preserves the existing state instead of minting a fresh provisioning_id.
+  state_action="$(python3 - "$STATE_DIR/state.json" "$DEALER_ID" <<'PY'
 import json, os, sys, tempfile, uuid
 path, dealer = sys.argv[1:]
-state={'phase':'PROVISIONED_OFFLINE','enrollment_status':'PENDING','fleet_status':'PENDING','device_id':'BF-'+dealer,'provisioning_id':str(uuid.uuid4())}
-fd,tmp=tempfile.mkstemp(prefix='state.json.',dir=os.path.dirname(path)); os.fchmod(fd,0o600)
-with os.fdopen(fd,'w') as f: json.dump(state,f,sort_keys=True); f.write('\n')
-os.replace(tmp,path)
+try:
+    with open(path, encoding='utf-8') as f:
+        existing = json.load(f)
+except (OSError, ValueError):
+    existing = {}
+phase = existing.get('phase')
+if phase in ('ENROLLED', 'READY'):
+    print('PRESERVED ' + phase)
+else:
+    state={'phase':'PROVISIONED_OFFLINE','enrollment_status':'PENDING','fleet_status':'PENDING','device_id':'BF-'+dealer,'provisioning_id':str(uuid.uuid4())}
+    fd,tmp=tempfile.mkstemp(prefix='state.json.',dir=os.path.dirname(path)); os.fchmod(fd,0o600)
+    with os.fdopen(fd,'w') as f: json.dump(state,f,sort_keys=True); f.write('\n')
+    os.replace(tmp,path)
+    print('PROVISIONED')
 PY
+)"
+  if [[ "$state_action" == PRESERVED* ]]; then
+    log "state preserved (phase=${state_action#PRESERVED })"
+  else
     chmod 600 "$STATE_DIR/state.json"
     log "OFFLINE COMPLETE: device is PROVISIONED_OFFLINE; enrollment is required before READY"
     notice "Kurulum tamam: $BF_DEVICE yerel paket deposundan kuruldu."
     notice 'Cihaz şu an PROVISIONED_OFFLINE durumunda; kayıt (enrollment) tamamlanana kadar READY sayılmaz.'
+  fi
 fi
 log "=== install done dealer=$BF_DEVICE failures=$failures ==="