better cache management

.github/workflows/vulnix.yml

@@ -5,6 +5,10 @@ on:
     - cron: "0 6 * * 1"
   workflow_dispatch:
 
+permissions:
+  contents: read
+  issues: write
+
 jobs:
   scan:
     runs-on: nix
@@ -21,7 +25,7 @@ jobs:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           API: ${{ github.api_url }}
           REPO: ${{ github.repository }}
-        run: 'nix run -L .#ci -- upsert-issue vulnix "vulnix: known vulnerabilities in system closure" "$RUNNER_TEMP/issue.md"'
+        run: nix run -L .#ci -- upsert-issue vulnix "vulnix: known vulnerabilities in system closure" "$RUNNER_TEMP/issue.md"
 
       - name: close resolved issue
         if: steps.scan.outputs.count == '0'

ci.nix

@@ -1,17 +1,10 @@
 { lib, pkgs }:
-let
-  # vulnix's bundled nix-2.34 broke its own --experimental-features detection
-  nixConf = pkgs.writeText "ci-nix.conf" "experimental-features = nix-command flakes";
-in
 # tools must track ci.py's subprocess calls; nix stays ambient (uses the
 # runner's own nix, not a pinned one).
 pkgs.writers.writePython3Bin "ci" {
   flakeIgnore = [ "E501" ];
   libraries = [ pkgs.python3Packages.subprocess-tee ];
   makeWrapperArgs = [
-    "--set-default"
-    "NIX_USER_CONF_FILES"
-    (toString nixConf)
     "--prefix"
     "PATH"
     ":"

ci.py

@@ -18,8 +18,7 @@ def run(*args, **kw):
 
 
 def nix(*args):
-    # tee to stderr - stdout leaks into GITHUB_OUTPUT $() captures
-    return subprocess_tee.run(("nix", "--extra-experimental-features", "nix-command flakes", "-L", *args), check=True, text=True, stdout=sys.stderr).stdout
+    return subprocess_tee.run(("nix", "--extra-experimental-features", "nix-command flakes", "-L", *args), check=True, text=True).stdout
 
 
 def git(*args, check=True):
@@ -33,16 +32,13 @@ def gitea(method, path, body=None):
     req.add_header("Authorization", f'token {os.environ["GH_TOKEN"]}')
     if data is not None:
         req.add_header("Content-Type", "application/json")
-    try:
     with urllib.request.urlopen(req) as resp:
         raw = resp.read()
-    except urllib.error.HTTPError as e:
-        sys.exit(f"gitea {method} {path}: {e.code} {e.reason}\n{e.read().decode()}")
     return json.loads(raw) if raw else None
 
 
 def find_open_issue(label):
-    issues = gitea("GET", f"/issues?state=open&type=issues&labels={label}")
+    issues = gitea("GET", f"/issues?state=open&labels={label}")
     return issues[0] if issues else None
 
 
@@ -239,7 +235,7 @@ def format_vulnix(vulns):
     for v in vulns:
         for cve in v["affected_by"]:
             score = v.get("cvssv3_basescore", {}).get(cve)
-            desc = v.get("description", {}).get(cve, "").replace("\n", " ").replace("|", "∣")
+            desc = v.get("description", {}).get(cve, "")
             if len(desc) > 120:
                 desc = desc[:117] + "..."
             rows.append((v["pname"], v["version"], cve, score, desc))

configuration.nix

@@ -4,27 +4,12 @@
   lib,
   hostInfo,
   customPkgs,
-  flakeRev,
   ...
 }:
 let
   inherit (hostInfo) paths;
 in
 {
-  imports = [ ./modules/autodeploy.nix ];
-
-  services.server-autodeploy = {
-    enable = true;
-    giteaUrl = "https://git.floweytf.com/";
-    giteaRepo = "server/server-config";
-    tokenFile = "${hostInfo.paths.secrets}/gitea-autodeploy-token";
-    allowedSigners = [
-      "flowey@arch namespaces=\"git\" ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIPaGpTX3EzfezVc2nl3MgsKVamYYP2h2VJaRny8pZkvi"
-    ];
-    stateDir = hostInfo.paths.config;
-    repoOwner = "server";
-  };
-
   assertions =
     let
       isMounted =
@@ -62,14 +47,6 @@ in
 
   time.timeZone = "UTC";
   system.stateVersion = "26.05";
-  system.nixos.label =
-    let
-      labelFile = ./.nixos-label;
-    in
-    if builtins.pathExists labelFile then
-      lib.removeSuffix "\n" (builtins.readFile labelFile)
-    else
-      flakeRev;
 
   nix = {
     settings = {

containers/server-n0/mail.nix

@@ -146,8 +146,6 @@ let
 in
 mkContainer {
   inherit name;
-  includedInTests = false; # TODO: fixme, clearly mail is flakey
-
   net.access = [ { to = "@internet"; } ];
   svc.autoStart = true;
   net.forward.tcp = [

containers/server-n0/ssh.nix

@@ -61,16 +61,12 @@ let
       net ? { },
       resources ? { },
       traits ? [ ],
-      isolation ? { },
     }:
     mkContainer {
       inherit name;
       inherit net resources;
       svc.autoStart = true;
-      isolation = {
-        inherit nixDaemon;
-      }
-      // isolation;
+      isolation.nixDaemon = nixDaemon;
       traits = [ sshTrait ] ++ traits;
       packages = with pkgs; [ tmux ] ++ packages;
       miscProperties = self._sshpiperRegister routingInfo;
@@ -117,28 +113,14 @@ in
   ssh-generic = mkSsh {
     name = "ssh-generic";
     net.access = [ { to = "@internet"; } ];
-    # /dev/kvm keeps its host ownership, so the kvm gid must not be shifted.
-    isolation.privateUsers = "identity";
     resources = {
       fs."/home/runner/data" = {
         hostPath = "${data}/ssh/generic";
         isReadOnly = false;
       };
-      devices = [ "/dev/kvm" ];
       mem = "16G";
       cpu = 65;
     };
-    packages = with pkgs; [ qemu_kvm ];
-    traits = [
-      {
-        containerConfig =
-          { config, ... }:
-          {
-            users.groups.kvm.gid = config.ids.gids.kvm;
-            users.users.runner.extraGroups = [ "kvm" ];
-          };
-      }
-    ];
     routingInfo = [
       {
         username = "autumn";

dsl.nix

@@ -221,17 +221,6 @@ rec {
     `resources.fs`
     : Bind mounts, keyed by mount point.
 
-    `resources.devices`
-    : Host device nodes to expose, as paths. Bound into the container at the
-      same path and added to the unit's `DeviceAllow=` - the cgroup
-      permission alone doesn't get you a working node (nothing can `mknod` a
-      device inside a user namespace), and the bind alone leaves the cgroup
-      denying `open()`, so both are needed. The bind is unconditional, so
-      the device must actually exist on whatever host runs the container.
-      Nodes keep their host ownership, so a consumer inside needs
-      `isolation.privateUsers = "identity"` for the owning gid to mean
-      anything.
-
     `isolation.nixDaemon`
     : Whether the Nix daemon and CLI are available inside the container.
 
@@ -281,7 +270,6 @@ rec {
       cpuQuota = "${toString (cpu * 100)}%";
       memoryMax = resources.mem or "512M";
       bindMounts = resources.fs or { };
-      devices = resources.devices or [ ];
       allowsNixDaemon = isolation.nixDaemon or false;
       privateUsers = isolation.privateUsers or "pick";
 
@@ -289,7 +277,7 @@ rec {
       allBindMounts = builtins.foldl' (acc: t: acc // (t.resources.fs or { })) { } traits // bindMounts;
       userModules = map mkUser (lib.concatMap (t: t.users or [ ]) traits ++ users);
       allPorts = ports ++ map (p: (normalizePort p).container) hostPorts;
-      useIdmap = privateUsers == "pick" || (builtins.isInt privateUsers && privateUsers > 0);
+      useIdmap = privateUsers != "no";
       bindMountExtraFlags = lib.mapAttrsToList (
         mountPoint: bm:
         (if bm.isReadOnly then "--bind-ro=" else "--bind=")
@@ -297,10 +285,6 @@ rec {
         + lib.optionalString useIdmap ":idmap"
       ) allBindMounts;
       bindMountHostPaths = map (bm: bm.hostPath) (lib.attrValues allBindMounts);
-      # DeviceAllow= alone only grants cgroup permission - nspawn can't
-      # mknod a device node inside a user namespace, so the node still
-      # needs an explicit bind from the host.
-      deviceFlags = map (d: "--bind=${d}") devices;
       inherit (deriveIdentity ifPrefix.nspawn name)
         iface
         mac
@@ -379,7 +363,6 @@ rec {
         hostAddr
         bindMountHostPaths
         includedInTests
-        devices
         ;
       bindMounts = allBindMounts;
       kind = "nspawn";
@@ -387,7 +370,7 @@ rec {
       nixosContainer = {
         inherit autoStart privateUsers;
         bindMounts = { };
-        extraFlags = bindMountExtraFlags ++ deviceFlags;
+        extraFlags = bindMountExtraFlags;
         privateNetwork = true;
         extraVeths = linkVeths;
         config =
@@ -520,7 +503,6 @@ rec {
         ;
       inherit needsGateway;
       bindMountHostPaths = map (s: s.source) shares;
-      devices = [ ];
       kind = "vm";
       miscProperties = flattenMiscProp self miscProperties;
       microvmConfig = {
@@ -957,7 +939,6 @@ rec {
           startLimitBurst = 5;
           startLimitIntervalSec = 120;
           unitConfig.RequiresMountsFor = c.bindMountHostPaths;
-          serviceConfig.DeviceAllow = map (d: "${d} rw") c.devices;
         };
 
       # the host end of a link, for both kinds. nixos-containers only configures
@@ -1317,7 +1298,6 @@ rec {
         HOST_ADDR = ${pyDict pyStr (lib.mapAttrs (_: c: c.hostAddr) autoStartTestContainers)}
         IFACE = ${pyDict pyStr (lib.mapAttrs (_: c: c.iface) autoStartTestContainers)}
         KIND = ${pyDict pyStr (lib.mapAttrs (_: c: c.kind) autoStartTestContainers)}
-        DEVICES = ${pyDict pyList (lib.mapAttrs (_: c: c.devices) autoStartTestContainers)}
         # only nspawn guests can be entered with nsenter; a vm has no host-side
         # netns, so every probe that runs *inside* a guest is limited to these
         enterable = [c for c in included if KIND[c] == "nspawn"]

flake.lock

@@ -26,11 +26,11 @@
         "spectrum": "spectrum"
       },
       "locked": {
-        "lastModified": 1789506833,
-        "narHash": "sha256-4L8SOCF84de1irB72Ka9VHcu9wrAJ9Amuey4lU9oh4Y=",
+        "lastModified": 1786300091,
+        "narHash": "sha256-4UFJOVGpaYtHW5yasSv80MaJxTBYdk2zyf3jhKtt0wA=",
         "owner": "microvm-nix",
         "repo": "microvm.nix",
-        "rev": "187b0a390ee054028106a674e7b01b1cb940cbba",
+        "rev": "71beea0076cd46dafcee97a5a2e7d00cbba5bd2f",
         "type": "github"
       },
       "original": {
@@ -63,11 +63,11 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1789546076,
-        "narHash": "sha256-zVxLZiSnmaaPLwnhj7pwmqe3axBg/C6nG5JZsJMh2g4=",
+        "lastModified": 1786384358,
+        "narHash": "sha256-RzPPiWeUtuvymnpuEWsdtzli5w4kjZs49FqEs3/1u+I=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "b1b875982b17dabde9b4a37f3e229e74913e6db3",
+        "rev": "2fcb964de67fcf60b43471c55d5d99e61a9ccb5a",
         "type": "github"
       },
       "original": {
@@ -87,11 +87,11 @@
     "spectrum": {
       "flake": false,
       "locked": {
-        "lastModified": 1786032750,
-        "narHash": "sha256-+Ba/kIn5BjvV7PN8bWOo7RoqgTY/Yv2M7qeMf3wS2Ag=",
+        "lastModified": 1785761586,
+        "narHash": "sha256-MWOMVqJJERwjQGgRIv8d6rlEBqbLM3VZWxHkNKIIZNw=",
         "ref": "refs/heads/main",
-        "rev": "77d2389874bb092b8dbf05ecc5b1030372e80e98",
-        "revCount": 1417,
+        "rev": "a7762d6f54b40560dd5255ce902e6e6a5d980fe9",
+        "revCount": 1416,
         "type": "git",
         "url": "https://spectrum-os.org/git/spectrum"
       },

flake.nix

@@ -11,7 +11,6 @@
 
   outputs =
     {
-      self,
       nixpkgs,
       mtce,
       microvm,
@@ -22,8 +21,6 @@
       inherit (nixpkgs) lib;
       pkgs = import nixpkgs { inherit system; };
 
-      flakeRev = self.shortRev or self.dirtyShortRev or "unknown";
-
       customPkgs = {
         hopper-rs = pkgs.callPackage ./packages/hopper-rs.nix { };
         sshpiper = pkgs.callPackage ./packages/sshpiper.nix { };
@@ -32,7 +29,6 @@
         server-rebuild = pkgs.callPackage ./packages/server-rebuild.nix { };
         server-status = pkgs.callPackage ./packages/server-status.nix { };
         server-kexec = pkgs.callPackage ./packages/server-kexec.nix { };
-        server-autodeploy = pkgs.callPackage ./packages/server-autodeploy.nix { };
       };
 
       ci = import ./ci.nix { inherit lib pkgs; };
@@ -50,13 +46,12 @@
           server = "/persist/@containers";
           data = "/persist/@containers/data";
           secrets = "/persist/@secrets/";
-          config = "/persist/@config/";
         };
       };
 
       specialArgsFor = hostName: {
         hostInfo = hostInfoFor hostName;
-        inherit inputs customPkgs flakeRev;
+        inherit inputs customPkgs;
       };
 
       hostDefFor =
@@ -152,7 +147,6 @@
           server-rebuild
           server-status
           server-kexec
-          server-autodeploy
           ;
       };
     };

modules/autodeploy.nix

@@ -1,96 +0,0 @@
-{
-  config,
-  lib,
-  pkgs,
-  customPkgs,
-  ...
-}:
-let
-  cfg = config.services.server-autodeploy;
-in
-{
-  options.services.server-autodeploy = {
-    enable = lib.mkEnableOption "automatic deployment from signed git tags";
-
-    interval = lib.mkOption {
-      type = lib.types.str;
-      default = "15min";
-    };
-
-    branch = lib.mkOption {
-      type = lib.types.str;
-      default = "master";
-    };
-
-    giteaUrl = lib.mkOption {
-      type = lib.types.str;
-    };
-
-    giteaRepo = lib.mkOption {
-      type = lib.types.str;
-    };
-
-    tokenFile = lib.mkOption {
-      type = lib.types.str;
-    };
-
-    allowedSigners = lib.mkOption {
-      type = lib.types.listOf lib.types.str;
-      default = [ ];
-    };
-    
-    repoOwner = lib.mkOption {
-      type = lib.types.str;
-      description = "User that owns the configuration checkout";
-    };
-
-    stateDir = lib.mkOption {
-      type = lib.types.str;
-    };
-
-    hostname = lib.mkOption {
-      type = lib.types.str;
-      default = config.networking.hostName;
-    };
-  };
-
-  config = lib.mkIf cfg.enable {
-    systemd.services.server-autodeploy = {
-      description = "Deploy latest signed tag";
-      after = [ "network-online.target" ];
-      wants = [ "network-online.target" ];
-
-      environment = {
-        GITEA_URL = cfg.giteaUrl;
-        GITEA_REPO = cfg.giteaRepo;
-        TOKEN_FILE = cfg.tokenFile;
-        ALLOWED_SIGNERS = builtins.toFile "signers" (lib.join "\n" cfg.allowedSigners);
-        DEPLOY_BRANCH = cfg.branch;
-        STATE_DIR = cfg.stateDir;
-        REPO_DIR = "/persist/@configuration";
-        REPO_OWNER = cfg.repoOwner;
-        HOSTNAME = cfg.hostname;
-      };
-
-      serviceConfig = {
-        Type = "oneshot";
-        ExecStart = lib.getExe customPkgs.server-autodeploy;
-        TimeoutStartSec = "30min";
-      };
-
-      path = [
-        config.nix.package
-        pkgs.nixos-rebuild
-      ];
-    };
-
-    systemd.timers.server-autodeploy = {
-      wantedBy = [ "timers.target" ];
-      timerConfig = {
-        OnBootSec = "2min";
-        OnUnitActiveSec = cfg.interval;
-        RandomizedDelaySec = "2min";
-      };
-    };
-  };
-}

packages/server-autodeploy.nix

@@ -1,18 +0,0 @@
-{
-  writeShellApplication,
-  git,
-  curl,
-  jq,
-  util-linux,
-  openssh,
-}:
-import ./mk-script.nix { inherit writeShellApplication; } {
-  name = "server-autodeploy";
-  runtimeInputs = [
-    git
-    curl
-    jq
-    util-linux
-    openssh
-  ];
-}

scripts/server-autodeploy.sh

@@ -1,60 +0,0 @@
-REPO_DIR="${REPO_DIR:-/persist/@configuration}"
-STATE_DIR="${STATE_DIR:-/persist/@autodeploy}"
-BRANCH="origin/${DEPLOY_BRANCH:-master}"
-
-mkdir -p "$STATE_DIR"
-exec 9>"$STATE_DIR/lock"
-flock -n 9 || { echo "another deploy running"; exit 0; }
-
-runas() { runuser -u "$REPO_OWNER" -- "$@"; }
-
-report_status() {
-  echo "updating status to: $1 - $2"
-  [[ -f "$TOKEN_FILE" ]] || return 0
-  curl -sf -X POST \
-    -H "Authorization: token $(<"$TOKEN_FILE")" \
-    -H "Content-Type: application/json" \
-    -d "$(jq -n --arg s "$1" --arg d "$2" --arg c "deploy/$HOSTNAME" \
-      '{state:$s,description:$d,context:$c}')" \
-    "$GITEA_URL/api/v1/repos/$GITEA_REPO/statuses/$3" >/dev/null
-}
-
-cd "$REPO_DIR"
-
-export GIT_CONFIG_COUNT=2
-export GIT_CONFIG_KEY_0=gpg.format
-export GIT_CONFIG_VALUE_0=ssh
-export GIT_CONFIG_KEY_1=gpg.ssh.allowedSignersFile
-export GIT_CONFIG_VALUE_1="$ALLOWED_SIGNERS"
-
-runas git fetch origin --tags --prune-tags
-
-latest=$(runas git describe --abbrev=0 "$BRANCH" 2>/dev/null) || {
-  echo "no tags on $BRANCH"
-  exit 0
-}
-
-current=$(cat "$STATE_DIR/tag" 2>/dev/null) || true
-[[ "$latest" != "$current" ]] || { echo "already at $latest"; exit 0; }
-
-echo "new tag: $latest (current: ${current:-none})"
-
-runas git tag -v "$latest" || {
-  sha=$(runas git rev-list -1 "$latest")
-  report_status "failure" "tag signature verification failed" "$sha"
-  exit 1
-}
-
-sha=$(runas git rev-list -1 "$latest")
-report_status "pending" "deploying $latest" "$sha"
-
-flake_url="git+${GITEA_URL}/${GITEA_REPO}?ref=refs/tags/${latest}&rev=${sha}#$HOSTNAME"
-
-if nixos-rebuild switch --flake "$flake_url"; then
-  echo "$latest" > "$STATE_DIR/tag"
-  report_status "success" "deployed $latest" "$sha"
-  echo "deployed $latest"
-else
-  report_status "failure" "nixos-rebuild failed" "$sha"
-  exit 1
-fi

tests/server-n0/integration.py

@@ -27,7 +27,6 @@ if TYPE_CHECKING:
     HOST_ADDR: dict[str, str] = {}
     IFACE: dict[str, str] = {}
     KIND: dict[str, str] = {}
-    DEVICES: dict[str, list[str]] = {}
     EDGES: list[tuple[str, str, list[int], list[int]]] = []
     SSH_KEY: str = ""
     SSH_OPTS: str = ""
@@ -56,11 +55,6 @@ def container_pid(name):
     return server.succeed(f"machinectl show {name} -p Leader --value").strip()
 
 
-def ct_cmd(name, cmd):
-    pid = container_pid(name)
-    return f"{NSENTER} --mount=/proc/{pid}/ns/mnt --root=/proc/{pid}/root -- {SH} -c '{cmd}'"
-
-
 def net_cmd(name, cmd):
     pid = container_pid(name)
     return f"{NSENTER} --net=/proc/{pid}/ns/net -- {cmd}"
@@ -153,8 +147,6 @@ with subtest("(functional): web services respond"):
     for name, port in [("devenv", 8080), ("maven", 8080), ("git", 3000)]:
         server.succeed(net_cmd(name, f"{CURL} -sf --max-time 5 -o /dev/null http://{ADDR[name]}:{port}/ >&2"))
 
-# TODO: re-enable mail
-"""
 with subtest("(functional): mail accepts connections"):
     smtp_probe = net_cmd("mail", f"{SH} -c 'exec 3<>/dev/tcp/{ADDR['mail']}/25 && head -1 <&3'")
     server.wait_until_succeeds(f"{smtp_probe} | grep -q '^220'", timeout=60)
@@ -169,7 +161,6 @@ with subtest("(functional): mail accepts connections"):
         return greeting.strip().startswith("* OK")
 
     retry(imap_ready, timeout_seconds=60)
-"""
 
 with subtest("(functional): nginx proxies to backend containers"):
     server.succeed(
@@ -200,7 +191,7 @@ with subtest("(isolation): undeclared container edges are refused"):
         ("devenv", "git", 3000),
         ("git", "devenv", 8080),
         ("maven", "git", 3000),
-        # ("nginx", "mail", 25),
+        ("nginx", "mail", 25),
         ("ssh-botm", "ssh-admin", 22),
         ("overseer", "git", 3000),
     ]:
@@ -253,7 +244,7 @@ with subtest("(isolation): renaming the interface changes nothing"):
     rename(iface, "evil0")
     try:
         server.succeed(ct_connect("nginx", ADDR["git"], 3000))
-        # server.fail(ct_connect("nginx", ADDR["mail"], 25, timeout=2))
+        server.fail(ct_connect("nginx", ADDR["mail"], 25, timeout=2))
     finally:
         rename("evil0", iface)
 
@@ -332,40 +323,3 @@ with subtest("(functional): published ports follow the container's lifetime"):
     server.succeed("systemctl start container@nginx")
     server.wait_for_unit("container@nginx.service")
     internet.wait_until_succeeds(ext_connect(80), timeout=30)
-
-with subtest("(functional): declared devices are usable in-container"):
-    for name, devices in DEVICES.items():
-        for dev in devices:
-            # opened, not stat'd: without DeviceAllow= the node is still visible
-            server.succeed(ct_cmd(name, f"exec 3<> {dev}"))
-
-
-with subtest("(functional): /dev/kvm actually initializes, not just opens"):
-    # a wrong-gid or unmapped node still opens but fails the ioctls qemu
-    # needs, so boot with no disk (never completes) and kill it - the
-    # failure mode this catches is qemu refusing to start at all.
-    for name, devices in DEVICES.items():
-        if "/dev/kvm" not in devices:
-            continue
-        out = server.succeed(
-            ct_cmd(
-                name,
-                f"{TIMEOUT} 3 qemu-system-x86_64 -enable-kvm -m 256 -nographic -no-reboot "
-                ">/tmp/qemu.out 2>&1; cat /tmp/qemu.out",
-            )
-        )
-        assert "Permission denied" not in out, f"{name}: qemu couldn't use /dev/kvm:\n{out}"
-        assert "Could not access KVM" not in out, f"{name}: qemu couldn't use /dev/kvm:\n{out}"
-
-
-with subtest("(isolation): devices reach only the container that declared them"):
-    for name in enterable:
-        for dev in {d for ds in DEVICES.values() for d in ds} - set(DEVICES.get(name, [])):
-            server.fail(ct_cmd(name, f"test -e {dev}"))
-
-
-with subtest("(isolation): identity uid mapping keeps the user namespace"):
-    for name in (n for n, ds in DEVICES.items() if ds):
-        host_ns = server.succeed("readlink /proc/1/ns/user").strip()
-        ct_ns = server.succeed(f"readlink /proc/{container_pid(name)}/ns/user").strip()
-        assert host_ns != ct_ns, f"{name} shares the host user namespace"