containerization (#3)

Co-authored-by: Floweynt <50057682+Floweynt@users.noreply.github.com>
Co-authored-by: Autumn Liu <al63244@my.utexas.edu>
Reviewed-on: #3
Co-authored-by: autumn <aut.tea7@gmail.com>

dsl.nix

@@ -221,6 +221,15 @@ rec {
     `resources.fs`
     : Bind mounts, keyed by mount point.
 
+    `resources.devices`
+    : Host device nodes to expose, as paths, added to the unit's
+      `DeviceAllow=`. nspawn populates a handful of privileged devices
+      (`/dev/kvm`, `/dev/net/tun`, ...) into the container's own `/dev` once
+      the cgroup allows it, and silently skips ones the host doesn't have -
+      no `resources.fs` bind mount needed. 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.
 
@@ -228,17 +237,6 @@ rec {
     : systemd-nspawn user namespacing; see `--private-users=` in
       systemd-nspawn(1). Bind mounts always go through idmapped `--bind`.
 
-    `isolation.allowedDevices`
-    : List of `{ node; modifier; }` host devices to grant the container's
-      cgroup access to; see `containers.<name>.allowedDevices` in the NixOS
-      manual. nspawn bind-mounts a handful of devices (`/dev/kvm`,
-      `/dev/net/tun`, ...) into the container itself once the cgroup allows it -
-      this doesn't need a matching `resources.fs` bind mount.
-
-    `isolation.additionalCapabilities`
-    : Extra Linux capabilities granted to the container; see
-      `containers.<name>.additionalCapabilities` in the NixOS manual.
-
     `miscProperties`
     : Arbitrary data attached to the result. Used for cross-container discovery
       via `collectContainers`.
@@ -281,16 +279,15 @@ 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";
-      allowedDevices = isolation.allowedDevices or [ ];
-      additionalCapabilities = isolation.additionalCapabilities or [ ];
 
       allowInternet = hasInternetAccess access;
       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 != "no";
+      useIdmap = privateUsers == "pick" || (builtins.isInt privateUsers && privateUsers > 0);
       bindMountExtraFlags = lib.mapAttrsToList (
         mountPoint: bm:
         (if bm.isReadOnly then "--bind-ro=" else "--bind=")
@@ -376,17 +373,13 @@ rec {
         hostAddr
         bindMountHostPaths
         includedInTests
+        devices
         ;
       bindMounts = allBindMounts;
       kind = "nspawn";
       miscProperties = flattenMiscProp self miscProperties;
       nixosContainer = {
-        inherit
-          autoStart
-          privateUsers
-          allowedDevices
-          additionalCapabilities
-          ;
+        inherit autoStart privateUsers;
         bindMounts = { };
         extraFlags = bindMountExtraFlags;
         privateNetwork = true;
@@ -521,6 +514,7 @@ rec {
         ;
       inherit needsGateway;
       bindMountHostPaths = map (s: s.source) shares;
+      devices = [ ];
       kind = "vm";
       miscProperties = flattenMiscProp self miscProperties;
       microvmConfig = {
@@ -957,6 +951,7 @@ 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
@@ -1316,6 +1311,7 @@ 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"]