Merge pull request #9928 from pendulm/fix_rootless_socket_activation

Fix rootless socket activation

2 files changed, 149 insertions, 12 deletions

diff --git a/pkg/rootless/rootless_linux.c b/pkg/rootless/rootless_linux.c
index 7a2bf0377..918b9a7e6 100644
--- a/pkg/rootless/rootless_linux.c
+++ b/pkg/rootless/rootless_linux.c
@@ -61,6 +61,10 @@ static int open_files_max_fd;
 static fd_set *open_files_set;
 static uid_t rootless_uid_init;
 static gid_t rootless_gid_init;
+static bool do_socket_activation = false;
+static char *saved_systemd_listen_fds;
+static char *saved_systemd_listen_pid;
+static char *saved_systemd_listen_fdnames;
 
 static int
 syscall_setresuid (uid_t ruid, uid_t euid, uid_t suid)
@@ -242,6 +246,10 @@ static void __attribute__((constructor)) init()
 {
   const char *xdg_runtime_dir;
   const char *pause;
+  const char *listen_pid;
+  const char *listen_fds;
+  const char *listen_fdnames;
+
   DIR *d;
 
   pause = getenv ("_PODMAN_PAUSE");
@@ -293,6 +301,26 @@ static void __attribute__((constructor)) init()
       closedir (d);
     }
 
+    listen_pid = getenv("LISTEN_PID");
+    listen_fds = getenv("LISTEN_FDS");
+    listen_fdnames = getenv("LISTEN_FDNAMES");
+
+    if (listen_pid != NULL && listen_fds != NULL && strtol(listen_pid, NULL, 10) == getpid())
+      {
+        // save systemd socket environment for rootless child
+        do_socket_activation = true;
+        saved_systemd_listen_pid = strdup(listen_pid);
+        saved_systemd_listen_fds = strdup(listen_fds);
+        saved_systemd_listen_fdnames = strdup(listen_fdnames);
+        if (saved_systemd_listen_pid == NULL
+                || saved_systemd_listen_fds == NULL
+                || saved_systemd_listen_fdnames == NULL)
+          {
+            fprintf (stderr, "save socket listen environments error: %s\n", strerror (errno));
+            _exit (EXIT_FAILURE);
+          }
+      }
+
   /* Shortcut.  If we are able to join the pause pid file, do it now so we don't
      need to re-exec.  */
   xdg_runtime_dir = getenv ("XDG_RUNTIME_DIR");
@@ -635,6 +663,12 @@ reexec_userns_join (int pid_to_join, char *pause_pid_file_path)
       for (f = 3; f <= open_files_max_fd; f++)
         if (is_fd_inherited (f))
           close (f);
+      if (do_socket_activation)
+        {
+          unsetenv ("LISTEN_PID");
+          unsetenv ("LISTEN_FDS");
+          unsetenv ("LISTEN_FDNAMES");
+        }
 
       return pid;
     }
@@ -660,6 +694,15 @@ reexec_userns_join (int pid_to_join, char *pause_pid_file_path)
       _exit (EXIT_FAILURE);
     }
 
+  if (do_socket_activation)
+    {
+      char s[32];
+      sprintf (s, "%d", getpid());
+      setenv ("LISTEN_PID", s, true);
+      setenv ("LISTEN_FDS", saved_systemd_listen_fds, true);
+      setenv ("LISTEN_FDNAMES", saved_systemd_listen_fdnames, true);
+    }
+
   setenv ("_CONTAINERS_USERNS_CONFIGURED", "init", 1);
   setenv ("_CONTAINERS_ROOTLESS_UID", uid, 1);
   setenv ("_CONTAINERS_ROOTLESS_GID", gid, 1);
@@ -777,9 +820,6 @@ reexec_in_user_namespace (int ready, char *pause_pid_file_path, char *file_to_re
   char **argv;
   char uid[16];
   char gid[16];
-  char *listen_fds = NULL;
-  char *listen_pid = NULL;
-  bool do_socket_activation = false;
   char *cwd = getcwd (NULL, 0);
   sigset_t sigset, oldsigset;
 
@@ -789,14 +829,6 @@ reexec_in_user_namespace (int ready, char *pause_pid_file_path, char *file_to_re
       _exit (EXIT_FAILURE);
     }
 
-  listen_pid = getenv("LISTEN_PID");
-  listen_fds = getenv("LISTEN_FDS");
-
-  if (listen_pid != NULL && listen_fds != NULL)
-    {
-      if (strtol(listen_pid, NULL, 10) == getpid())
-        do_socket_activation = true;
-    }
 
   sprintf (uid, "%d", geteuid ());
   sprintf (gid, "%d", getegid ());
@@ -814,7 +846,7 @@ reexec_in_user_namespace (int ready, char *pause_pid_file_path, char *file_to_re
         {
           long num_fds;
 
-          num_fds = strtol (listen_fds, NULL, 10);
+          num_fds = strtol (saved_systemd_listen_fds, NULL, 10);
           if (num_fds != LONG_MIN && num_fds != LONG_MAX)
             {
               int f;
@@ -863,6 +895,8 @@ reexec_in_user_namespace (int ready, char *pause_pid_file_path, char *file_to_re
       char s[32];
       sprintf (s, "%d", getpid());
       setenv ("LISTEN_PID", s, true);
+      setenv ("LISTEN_FDS", saved_systemd_listen_fds, true);
+      setenv ("LISTEN_FDNAMES", saved_systemd_listen_fdnames, true);
     }
 
   setenv ("_CONTAINERS_USERNS_CONFIGURED", "init", 1);
diff --git a/test/system/270-socket-activation.bats b/test/system/270-socket-activation.bats
new file mode 100644
index 000000000..25206c6a7
--- /dev/null
+++ b/test/system/270-socket-activation.bats
@@ -0,0 +1,103 @@
+#!/usr/bin/env bats   -*- bats -*-
+#
+# Tests podman system service under systemd socket activation
+#
+
+load helpers
+
+SERVICE_NAME="podman_test_$(random_string)"
+
+SYSTEMCTL="systemctl"
+UNIT_DIR="/usr/lib/systemd/system"
+SERVICE_SOCK_ADDR="/run/podman/podman.sock"
+
+if is_rootless; then
+    UNIT_DIR="$HOME/.config/systemd/user"
+    mkdir -p $UNIT_DIR
+
+    SYSTEMCTL="$SYSTEMCTL --user"
+    if [ -z "$XDG_RUNTIME_DIR" ]; then
+        export XDG_RUNTIME_DIR=/run/user/$(id -u)
+    fi
+    SERVICE_SOCK_ADDR="$XDG_RUNTIME_DIR/podman/podman.sock"
+fi
+
+SERVICE_FILE="$UNIT_DIR/$SERVICE_NAME.service"
+SOCKET_FILE="$UNIT_DIR/$SERVICE_NAME.socket"
+
+
+function setup() {
+    skip_if_remote "systemd tests are meaningless over remote"
+
+    basic_setup
+
+    cat > $SERVICE_FILE <<EOF
+[Unit]
+Description=Podman API Service
+Requires=podman.socket
+After=podman.socket
+Documentation=man:podman-system-service(1)
+StartLimitIntervalSec=0
+
+[Service]
+Type=exec
+KillMode=process
+Environment=LOGGING="--log-level=info"
+ExecStart=$PODMAN $LOGGING system service -t 2
+EOF
+    cat > $SOCKET_FILE <<EOF
+[Unit]
+Description=Podman API Socket
+Documentation=man:podman-system-service(1)
+
+[Socket]
+ListenStream=%t/podman/podman.sock
+SocketMode=0660
+
+[Install]
+WantedBy=sockets.target
+EOF
+
+    # ensure pause die before each test runs
+    if is_rootless; then
+        local pause_pid="$XDG_RUNTIME_DIR/libpod/tmp/pause.pid"
+        if [ -f $pause_pid ]; then
+            kill -9 $(cat $pause_pid) 2> /dev/null
+            rm -f $pause_pid
+        fi
+    fi
+    $SYSTEMCTL start "$SERVICE_NAME.socket"
+}
+
+function teardown() {
+    $SYSTEMCTL stop "$SERVICE_NAME.socket"
+    rm -f "$SERVICE_FILE" "$SOCKET_FILE"
+    $SYSTEMCTL daemon-reload
+    basic_teardown
+}
+
+@test "podman system service - socket activation - no container" {
+    run curl -s --max-time 3 --unix-socket $SERVICE_SOCK_ADDR http://podman/libpod/_ping
+    is "$output" "OK" "podman service responses normally"
+}
+
+@test "podman system service - socket activation - exist container " {
+    run_podman run $IMAGE sleep 90
+    run curl -s --max-time 3 --unix-socket $SERVICE_SOCK_ADDR http://podman/libpod/_ping
+    is "$output" "OK" "podman service responses normally"
+}
+
+@test "podman system service - socket activation - kill rootless pause " {
+    if ! is_rootless; then
+        skip "root podman no need pause process"
+    fi
+    run_podman run $IMAGE sleep 90
+    local pause_pid="$XDG_RUNTIME_DIR/libpod/tmp/pause.pid"
+    if [ -f $pause_pid ]; then
+        kill -9 $(cat $pause_pid) 2> /dev/null
+    fi
+    run curl -s --max-time 3 --unix-socket $SERVICE_SOCK_ADDR http://podman/libpod/_ping
+    is "$output" "OK" "podman service responses normally"
+}
+
+# vim: filetype=sh