Task Control Groups: add procfs interface
authorPaul Menage <menage@google.com>
Fri, 19 Oct 2007 06:39:35 +0000 (23:39 -0700)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Fri, 19 Oct 2007 18:53:36 +0000 (11:53 -0700)
Add:

/proc/cgroups - general system info

/proc/*/cgroup - per-task cgroup membership info

[a.p.zijlstra@chello.nl: cgroups: bdi init hooks]
Signed-off-by: Paul Menage <menage@google.com>
Cc: Serge E. Hallyn <serue@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/base.c
include/linux/cgroup.h
kernel/cgroup.c

index 4fe74d156416f0936b686d57dc6ee7bd2518dba5..0e9a9aa9df64de7959d27b0d7ca4d7b426a957b5 100644 (file)
@@ -67,6 +67,7 @@
 #include <linux/mount.h>
 #include <linux/security.h>
 #include <linux/ptrace.h>
+#include <linux/cgroup.h>
 #include <linux/cpuset.h>
 #include <linux/audit.h>
 #include <linux/poll.h>
@@ -2132,6 +2133,9 @@ static const struct pid_entry tgid_base_stuff[] = {
 #endif
 #ifdef CONFIG_CPUSETS
        REG("cpuset",     S_IRUGO, cpuset),
+#endif
+#ifdef CONFIG_CGROUPS
+       REG("cgroup",  S_IRUGO, cgroup),
 #endif
        INF("oom_score",  S_IRUGO, oom_score),
        REG("oom_adj",    S_IRUGO|S_IWUSR, oom_adjust),
@@ -2418,6 +2422,9 @@ static const struct pid_entry tid_base_stuff[] = {
 #endif
 #ifdef CONFIG_CPUSETS
        REG("cpuset",    S_IRUGO, cpuset),
+#endif
+#ifdef CONFIG_CGROUPS
+       REG("cgroup",  S_IRUGO, cgroup),
 #endif
        INF("oom_score", S_IRUGO, oom_score),
        REG("oom_adj",   S_IRUGO|S_IWUSR, oom_adjust),
index b21cf093ac62af42a2a62c19d82822d8fc1c1ad1..a9553568118fed713c32bf3256fdecafd0d8b603 100644 (file)
@@ -29,6 +29,8 @@ extern void cgroup_fork(struct task_struct *p);
 extern void cgroup_fork_callbacks(struct task_struct *p);
 extern void cgroup_exit(struct task_struct *p, int run_callbacks);
 
+extern struct file_operations proc_cgroup_operations;
+
 /* Per-subsystem/per-cgroup state maintained by the system. */
 struct cgroup_subsys_state {
        /* The cgroup that this subsystem is attached to. Useful
index cc68fe68a60ebdfd4a07dc4d5fa955978b14de73..db245f19eb8a761d727da30fe1790f2a4c050a57 100644 (file)
@@ -33,6 +33,7 @@
 #include <linux/mutex.h>
 #include <linux/mount.h>
 #include <linux/pagemap.h>
+#include <linux/proc_fs.h>
 #include <linux/rcupdate.h>
 #include <linux/sched.h>
 #include <linux/seq_file.h>
@@ -247,13 +248,15 @@ static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, int mode);
 static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
 static int cgroup_populate_dir(struct cgroup *cont);
 static struct inode_operations cgroup_dir_inode_operations;
+static struct file_operations proc_cgroupstats_operations;
+
+static struct backing_dev_info cgroup_backing_dev_info = {
+       .capabilities   = BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_WRITEBACK,
+};
 
 static struct inode *cgroup_new_inode(mode_t mode, struct super_block *sb)
 {
        struct inode *inode = new_inode(sb);
-       static struct backing_dev_info cgroup_backing_dev_info = {
-               .capabilities   = BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_WRITEBACK,
-       };
 
        if (inode) {
                inode->i_mode = mode;
@@ -1600,6 +1603,11 @@ int __init cgroup_init(void)
 {
        int err;
        int i;
+       struct proc_dir_entry *entry;
+
+       err = bdi_init(&cgroup_backing_dev_info);
+       if (err)
+               return err;
 
        for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
                struct cgroup_subsys *ss = subsys[i];
@@ -1611,10 +1619,142 @@ int __init cgroup_init(void)
        if (err < 0)
                goto out;
 
+       entry = create_proc_entry("cgroups", 0, NULL);
+       if (entry)
+               entry->proc_fops = &proc_cgroupstats_operations;
+
 out:
+       if (err)
+               bdi_destroy(&cgroup_backing_dev_info);
+
        return err;
 }
 
+/*
+ * proc_cgroup_show()
+ *  - Print task's cgroup paths into seq_file, one line for each hierarchy
+ *  - Used for /proc/<pid>/cgroup.
+ *  - No need to task_lock(tsk) on this tsk->cgroup reference, as it
+ *    doesn't really matter if tsk->cgroup changes after we read it,
+ *    and we take cgroup_mutex, keeping attach_task() from changing it
+ *    anyway.  No need to check that tsk->cgroup != NULL, thanks to
+ *    the_top_cgroup_hack in cgroup_exit(), which sets an exiting tasks
+ *    cgroup to top_cgroup.
+ */
+
+/* TODO: Use a proper seq_file iterator */
+static int proc_cgroup_show(struct seq_file *m, void *v)
+{
+       struct pid *pid;
+       struct task_struct *tsk;
+       char *buf;
+       int retval;
+       struct cgroupfs_root *root;
+
+       retval = -ENOMEM;
+       buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
+       if (!buf)
+               goto out;
+
+       retval = -ESRCH;
+       pid = m->private;
+       tsk = get_pid_task(pid, PIDTYPE_PID);
+       if (!tsk)
+               goto out_free;
+
+       retval = 0;
+
+       mutex_lock(&cgroup_mutex);
+
+       for_each_root(root) {
+               struct cgroup_subsys *ss;
+               struct cgroup *cont;
+               int subsys_id;
+               int count = 0;
+
+               /* Skip this hierarchy if it has no active subsystems */
+               if (!root->actual_subsys_bits)
+                       continue;
+               for_each_subsys(root, ss)
+                       seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
+               seq_putc(m, ':');
+               get_first_subsys(&root->top_cgroup, NULL, &subsys_id);
+               cont = task_cgroup(tsk, subsys_id);
+               retval = cgroup_path(cont, buf, PAGE_SIZE);
+               if (retval < 0)
+                       goto out_unlock;
+               seq_puts(m, buf);
+               seq_putc(m, '\n');
+       }
+
+out_unlock:
+       mutex_unlock(&cgroup_mutex);
+       put_task_struct(tsk);
+out_free:
+       kfree(buf);
+out:
+       return retval;
+}
+
+static int cgroup_open(struct inode *inode, struct file *file)
+{
+       struct pid *pid = PROC_I(inode)->pid;
+       return single_open(file, proc_cgroup_show, pid);
+}
+
+struct file_operations proc_cgroup_operations = {
+       .open           = cgroup_open,
+       .read           = seq_read,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+
+/* Display information about each subsystem and each hierarchy */
+static int proc_cgroupstats_show(struct seq_file *m, void *v)
+{
+       int i;
+       struct cgroupfs_root *root;
+
+       mutex_lock(&cgroup_mutex);
+       seq_puts(m, "Hierarchies:\n");
+       for_each_root(root) {
+               struct cgroup_subsys *ss;
+               int first = 1;
+               seq_printf(m, "%p: bits=%lx cgroups=%d (", root,
+                          root->subsys_bits, root->number_of_cgroups);
+               for_each_subsys(root, ss) {
+                       seq_printf(m, "%s%s", first ? "" : ", ", ss->name);
+                       first = false;
+               }
+               seq_putc(m, ')');
+               if (root->sb) {
+                       seq_printf(m, " s_active=%d",
+                                  atomic_read(&root->sb->s_active));
+               }
+               seq_putc(m, '\n');
+       }
+       seq_puts(m, "Subsystems:\n");
+       for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
+               struct cgroup_subsys *ss = subsys[i];
+               seq_printf(m, "%d: name=%s hierarchy=%p\n",
+                          i, ss->name, ss->root);
+       }
+       mutex_unlock(&cgroup_mutex);
+       return 0;
+}
+
+static int cgroupstats_open(struct inode *inode, struct file *file)
+{
+       return single_open(file, proc_cgroupstats_show, 0);
+}
+
+static struct file_operations proc_cgroupstats_operations = {
+       .open = cgroupstats_open,
+       .read = seq_read,
+       .llseek = seq_lseek,
+       .release = single_release,
+};
+
 /**
  * cgroup_fork - attach newly forked task to its parents cgroup.
  * @tsk: pointer to task_struct of forking parent process.