drm: bridge: dw-hdmi-i2s-audio: enable INSERT_PCUV bit for LPCM
[firefly-linux-kernel-4.4.55.git] / kernel / cgroup.c
index 47851dbc893d1ee1e952bf6ef37792c59f1af77b..03a1b3f754d6aa3fc3be4aa3226beda3973bc3f6 100644 (file)
@@ -211,6 +211,7 @@ static unsigned long have_free_callback __read_mostly;
 /* Ditto for the can_fork callback. */
 static unsigned long have_canfork_callback __read_mostly;
 
+static struct file_system_type cgroup2_fs_type;
 static struct cftype cgroup_dfl_base_files[];
 static struct cftype cgroup_legacy_base_files[];
 
@@ -236,6 +237,9 @@ static int cgroup_addrm_files(struct cgroup_subsys_state *css,
  */
 static bool cgroup_ssid_enabled(int ssid)
 {
+       if (CGROUP_SUBSYS_COUNT == 0)
+               return false;
+
        return static_key_enabled(cgroup_subsys_enabled_key[ssid]);
 }
 
@@ -1647,10 +1651,6 @@ static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
                        all_ss = true;
                        continue;
                }
-               if (!strcmp(token, "__DEVEL__sane_behavior")) {
-                       opts->flags |= CGRP_ROOT_SANE_BEHAVIOR;
-                       continue;
-               }
                if (!strcmp(token, "noprefix")) {
                        opts->flags |= CGRP_ROOT_NOPREFIX;
                        continue;
@@ -1717,15 +1717,6 @@ static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
                        return -ENOENT;
        }
 
-       if (opts->flags & CGRP_ROOT_SANE_BEHAVIOR) {
-               pr_warn("sane_behavior: this is still under development and its behaviors will change, proceed at your own risk\n");
-               if (nr_opts != 1) {
-                       pr_err("sane_behavior: no other mount options allowed\n");
-                       return -EINVAL;
-               }
-               return 0;
-       }
-
        /*
         * If the 'all' option was specified select all the subsystems,
         * otherwise if 'none', 'name=' and a subsystem name options were
@@ -2004,6 +1995,7 @@ static struct dentry *cgroup_mount(struct file_system_type *fs_type,
                         int flags, const char *unused_dev_name,
                         void *data)
 {
+       bool is_v2 = fs_type == &cgroup2_fs_type;
        struct super_block *pinned_sb = NULL;
        struct cgroup_subsys *ss;
        struct cgroup_root *root;
@@ -2020,6 +2012,17 @@ static struct dentry *cgroup_mount(struct file_system_type *fs_type,
        if (!use_task_css_set_links)
                cgroup_enable_task_cg_lists();
 
+       if (is_v2) {
+               if (data) {
+                       pr_err("cgroup2: unknown option \"%s\"\n", (char *)data);
+                       return ERR_PTR(-EINVAL);
+               }
+               cgrp_dfl_root_visible = true;
+               root = &cgrp_dfl_root;
+               cgroup_get(&root->cgrp);
+               goto out_mount;
+       }
+
        mutex_lock(&cgroup_mutex);
 
        /* First find the desired set of subsystems */
@@ -2027,15 +2030,6 @@ static struct dentry *cgroup_mount(struct file_system_type *fs_type,
        if (ret)
                goto out_unlock;
 
-       /* look for a matching existing root */
-       if (opts.flags & CGRP_ROOT_SANE_BEHAVIOR) {
-               cgrp_dfl_root_visible = true;
-               root = &cgrp_dfl_root;
-               cgroup_get(&root->cgrp);
-               ret = 0;
-               goto out_unlock;
-       }
-
        /*
         * Destruction of cgroup root is asynchronous, so subsystems may
         * still be dying after the previous unmount.  Let's drain the
@@ -2146,9 +2140,10 @@ out_free:
 
        if (ret)
                return ERR_PTR(ret);
-
+out_mount:
        dentry = kernfs_mount(fs_type, flags, root->kf_root,
-                               CGROUP_SUPER_MAGIC, &new_sb);
+                             is_v2 ? CGROUP2_SUPER_MAGIC : CGROUP_SUPER_MAGIC,
+                             &new_sb);
        if (IS_ERR(dentry) || !new_sb)
                cgroup_put(&root->cgrp);
 
@@ -2191,6 +2186,12 @@ static struct file_system_type cgroup_fs_type = {
        .kill_sb = cgroup_kill_sb,
 };
 
+static struct file_system_type cgroup2_fs_type = {
+       .name = "cgroup2",
+       .mount = cgroup_mount,
+       .kill_sb = cgroup_kill_sb,
+};
+
 /**
  * task_cgroup_path - cgroup path of a task in the first cgroup hierarchy
  * @task: target task
@@ -2671,45 +2672,6 @@ static int cgroup_attach_task(struct cgroup *dst_cgrp,
        return ret;
 }
 
-int subsys_cgroup_allow_attach(struct cgroup_taskset *tset)
-{
-       const struct cred *cred = current_cred(), *tcred;
-       struct task_struct *task;
-       struct cgroup_subsys_state *css;
-
-       if (capable(CAP_SYS_NICE))
-               return 0;
-
-       cgroup_taskset_for_each(task, css, tset) {
-               tcred = __task_cred(task);
-
-               if (current != task && !uid_eq(cred->euid, tcred->uid) &&
-                   !uid_eq(cred->euid, tcred->suid))
-                       return -EACCES;
-       }
-
-       return 0;
-}
-
-static int cgroup_allow_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
-{
-       struct cgroup_subsys_state *css;
-       int i;
-       int ret;
-
-       for_each_css(css, i, cgrp) {
-               if (css->ss->allow_attach) {
-                       ret = css->ss->allow_attach(tset);
-                       if (ret)
-                               return ret;
-               } else {
-                       return -EACCES;
-               }
-       }
-
-       return 0;
-}
-
 static int cgroup_procs_write_permission(struct task_struct *task,
                                         struct cgroup *dst_cgrp,
                                         struct kernfs_open_file *of)
@@ -2724,24 +2686,9 @@ static int cgroup_procs_write_permission(struct task_struct *task,
         */
        if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
            !uid_eq(cred->euid, tcred->uid) &&
-           !uid_eq(cred->euid, tcred->suid)) {
-               /*
-                * if the default permission check fails, give each
-                * cgroup a chance to extend the permission check
-                */
-               struct cgroup_taskset tset = {
-                       .src_csets = LIST_HEAD_INIT(tset.src_csets),
-                       .dst_csets = LIST_HEAD_INIT(tset.dst_csets),
-                       .csets = &tset.src_csets,
-               };
-               struct css_set *cset;
-               cset = task_css_set(task);
-               list_add(&cset->mg_node, &tset.src_csets);
-               ret = cgroup_allow_attach(dst_cgrp, &tset);
-               list_del(&tset.src_csets);
-               if (ret)
-                       ret = -EACCES;
-       }
+           !uid_eq(cred->euid, tcred->suid) &&
+           !ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
+               ret = -EACCES;
 
        if (!ret && cgroup_on_dfl(dst_cgrp)) {
                struct super_block *sb = of->file->f_path.dentry->d_sb;
@@ -4848,6 +4795,7 @@ static void init_and_link_css(struct cgroup_subsys_state *css,
        memset(css, 0, sizeof(*css));
        css->cgroup = cgrp;
        css->ss = ss;
+       css->id = -1;
        INIT_LIST_HEAD(&css->sibling);
        INIT_LIST_HEAD(&css->children);
        css->serial_nr = css_serial_nr_next++;
@@ -5379,6 +5327,12 @@ int __init cgroup_init(void)
        BUG_ON(cgroup_init_cftypes(NULL, cgroup_dfl_base_files));
        BUG_ON(cgroup_init_cftypes(NULL, cgroup_legacy_base_files));
 
+       /*
+        * The latency of the synchronize_sched() is too high for cgroups,
+        * avoid it at the cost of forcing all readers into the slow path.
+        */
+       rcu_sync_enter_start(&cgroup_threadgroup_rwsem.rss);
+
        mutex_lock(&cgroup_mutex);
 
        /* Add init_css_set to the hash table */
@@ -5434,6 +5388,7 @@ int __init cgroup_init(void)
 
        WARN_ON(sysfs_create_mount_point(fs_kobj, "cgroup"));
        WARN_ON(register_filesystem(&cgroup_fs_type));
+       WARN_ON(register_filesystem(&cgroup2_fs_type));
        WARN_ON(!proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations));
 
        return 0;