core: general cgroup rework

Replace the very generic cgroup hookup with a much simpler one. With
this change only the high-level cgroup settings remain, the ability to
set arbitrary cgroup attributes is removed, so is support for adding
units to arbitrary cgroup controllers or setting arbitrary paths for
them (especially paths that are different for the various controllers).

This also introduces a new -.slice root slice, that is the parent of
system.slice and friends. This enables easy admin configuration of
root-level cgrouo properties.

This replaces DeviceDeny= by DevicePolicy=, and implicitly adds in
/dev/null, /dev/zero and friends if DeviceAllow= is used (unless this is
turned off by DevicePolicy=).

1 files changed, 11 insertions, 59 deletions

diff --git a/src/core/execute.c b/src/core/execute.c
index 9148d06df4..5e342f8d47 100644
--- a/src/core/execute.c
+++ b/src/core/execute.c
@@ -55,7 +55,6 @@
 #include "sd-messages.h"
 #include "ioprio.h"
 #include "securebits.h"
-#include "cgroup.h"
 #include "namespace.h"
 #include "tcpwrap.h"
 #include "exit-status.h"
@@ -67,6 +66,7 @@
 #include "syscall-list.h"
 #include "env-util.h"
 #include "fileio.h"
+#include "unit.h"
 
 #define IDLE_TIMEOUT_USEC (5*USEC_PER_SEC)
 
@@ -986,18 +986,17 @@ int exec_spawn(ExecCommand *command,
                bool apply_chroot,
                bool apply_tty_stdin,
                bool confirm_spawn,
-               CGroupBonding *cgroup_bondings,
-               CGroupAttribute *cgroup_attributes,
-               const char *cgroup_suffix,
+               CGroupControllerMask cgroup_mask,
+               const char *cgroup_path,
                const char *unit_id,
                int idle_pipe[2],
                pid_t *ret) {
 
+        _cleanup_strv_free_ char **files_env = NULL;
+        int socket_fd;
+        char *line;
         pid_t pid;
         int r;
-        char *line;
-        int socket_fd;
-        _cleanup_strv_free_ char **files_env = NULL;
 
         assert(command);
         assert(context);
@@ -1042,17 +1041,6 @@ int exec_spawn(ExecCommand *command,
                         NULL);
         free(line);
 
-        r = cgroup_bonding_realize_list(cgroup_bondings);
-        if (r < 0)
-                return r;
-
-        /* We must initialize the attributes in the parent, before we
-        fork, because we really need them initialized before making
-        the process a member of the group (which we do in both the
-        child and the parent), and we cannot really apply them twice
-        (due to 'append' style attributes) */
-        cgroup_attribute_apply_list(cgroup_attributes, cgroup_bondings);
-
         if (context->private_tmp && !context->tmp_dir && !context->var_tmp_dir) {
                 r = setup_tmpdirs(&context->tmp_dir, &context->var_tmp_dir);
                 if (r < 0)
@@ -1072,7 +1060,6 @@ int exec_spawn(ExecCommand *command,
                 _cleanup_strv_free_ char **our_env = NULL, **pam_env = NULL,
                         **final_env = NULL, **final_argv = NULL;
                 unsigned n_env = 0;
-                bool set_access = false;
 
                 /* child */
 
@@ -1185,8 +1172,8 @@ int exec_spawn(ExecCommand *command,
                         goto fail_child;
                 }
 
-                if (cgroup_bondings) {
-                        err = cgroup_bonding_install_list(cgroup_bondings, 0, cgroup_suffix);
+                if (cgroup_path) {
+                        err = cg_attach_with_mask(cgroup_mask, cgroup_path, 0);
                         if (err < 0) {
                                 r = EXIT_CGROUP;
                                 goto fail_child;
@@ -1269,36 +1256,6 @@ int exec_spawn(ExecCommand *command,
                                         goto fail_child;
                                 }
                         }
-
-                        if (cgroup_bondings && context->control_group_modify) {
-                                err = cgroup_bonding_set_group_access_list(cgroup_bondings, 0755, uid, gid);
-                                if (err >= 0)
-                                        err = cgroup_bonding_set_task_access_list(
-                                                        cgroup_bondings,
-                                                        0644,
-                                                        uid,
-                                                        gid,
-                                                        context->control_group_persistent);
-                                if (err < 0) {
-                                        r = EXIT_CGROUP;
-                                        goto fail_child;
-                                }
-
-                                set_access = true;
-                        }
-                }
-
-                if (cgroup_bondings && !set_access && context->control_group_persistent >= 0)  {
-                        err = cgroup_bonding_set_task_access_list(
-                                        cgroup_bondings,
-                                        (mode_t) -1,
-                                        (uid_t) -1,
-                                        (uid_t) -1,
-                                        context->control_group_persistent);
-                        if (err < 0) {
-                                r = EXIT_CGROUP;
-                                goto fail_child;
-                        }
                 }
 
                 if (apply_permissions) {
@@ -1562,7 +1519,8 @@ int exec_spawn(ExecCommand *command,
          * outside of the cgroup) and in the parent (so that we can be
          * sure that when we kill the cgroup the process will be
          * killed too). */
-        cgroup_bonding_install_list(cgroup_bondings, pid, cgroup_suffix);
+        if (cgroup_path)
+                cg_attach(SYSTEMD_CGROUP_CONTROLLER, cgroup_path, pid);
 
         exec_status_start(&command->exec_status, pid);
 
@@ -1578,7 +1536,6 @@ void exec_context_init(ExecContext *c) {
         c->cpu_sched_policy = SCHED_OTHER;
         c->syslog_priority = LOG_DAEMON|LOG_INFO;
         c->syslog_level_prefix = true;
-        c->control_group_persistent = -1;
         c->ignore_sigpipe = true;
         c->timer_slack_nsec = (nsec_t) -1;
 }
@@ -1843,8 +1800,7 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
         assert(c);
         assert(f);
 
-        if (!prefix)
-                prefix = "";
+        prefix = strempty(prefix);
 
         fprintf(f,
                 "%sUMask: %04o\n"
@@ -1852,8 +1808,6 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
                 "%sRootDirectory: %s\n"
                 "%sNonBlocking: %s\n"
                 "%sPrivateTmp: %s\n"
-                "%sControlGroupModify: %s\n"
-                "%sControlGroupPersistent: %s\n"
                 "%sPrivateNetwork: %s\n"
                 "%sIgnoreSIGPIPE: %s\n",
                 prefix, c->umask,
@@ -1861,8 +1815,6 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
                 prefix, c->root_directory ? c->root_directory : "/",
                 prefix, yes_no(c->non_blocking),
                 prefix, yes_no(c->private_tmp),
-                prefix, yes_no(c->control_group_modify),
-                prefix, yes_no(c->control_group_persistent),
                 prefix, yes_no(c->private_network),
                 prefix, yes_no(c->ignore_sigpipe));