use clamp_t in UNAME26 fix
[linux-2.6.git] / kernel / pid_namespace.c
index 86b3796..57bc1fd 100644 (file)
@@ -13,6 +13,9 @@
 #include <linux/syscalls.h>
 #include <linux/err.h>
 #include <linux/acct.h>
+#include <linux/slab.h>
+#include <linux/proc_fs.h>
+#include <linux/reboot.h>
 
 #define BITS_PER_PAGE          (PAGE_SIZE*8)
 
@@ -71,7 +74,7 @@ static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_p
 {
        struct pid_namespace *ns;
        unsigned int level = parent_pid_ns->level + 1;
-       int i;
+       int i, err = -ENOMEM;
 
        ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
        if (ns == NULL)
@@ -95,14 +98,20 @@ static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_p
        for (i = 1; i < PIDMAP_ENTRIES; i++)
                atomic_set(&ns->pidmap[i].nr_free, BITS_PER_PAGE);
 
+       err = pid_ns_prepare_proc(ns);
+       if (err)
+               goto out_put_parent_pid_ns;
+
        return ns;
 
+out_put_parent_pid_ns:
+       put_pid_ns(parent_pid_ns);
 out_free_map:
        kfree(ns->pidmap[0].page);
 out_free:
        kmem_cache_free(pid_ns_cachep, ns);
 out:
-       return ERR_PTR(-ENOMEM);
+       return ERR_PTR(err);
 }
 
 static void destroy_pid_namespace(struct pid_namespace *ns)
@@ -160,14 +169,9 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
        while (nr > 0) {
                rcu_read_lock();
 
-               /*
-                * Use force_sig() since it clears SIGNAL_UNKILLABLE ensuring
-                * any nested-container's init processes don't ignore the
-                * signal
-                */
                task = pid_task(find_vpid(nr), PIDTYPE_PID);
-               if (task)
-                       force_sig(SIGKILL, task);
+               if (task && !__fatal_signal_pending(task))
+                       send_sig_info(SIGKILL, SEND_SIG_FORCED, task);
 
                rcu_read_unlock();
 
@@ -180,13 +184,76 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
                rc = sys_wait4(-1, NULL, __WALL, NULL);
        } while (rc != -ECHILD);
 
+       if (pid_ns->reboot)
+               current->signal->group_exit_code = pid_ns->reboot;
+
        acct_exit_ns(pid_ns);
        return;
 }
 
+static int pid_ns_ctl_handler(struct ctl_table *table, int write,
+               void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+       struct ctl_table tmp = *table;
+
+       if (write && !capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       /*
+        * Writing directly to ns' last_pid field is OK, since this field
+        * is volatile in a living namespace anyway and a code writing to
+        * it should synchronize its usage with external means.
+        */
+
+       tmp.data = &current->nsproxy->pid_ns->last_pid;
+       return proc_dointvec(&tmp, write, buffer, lenp, ppos);
+}
+
+static struct ctl_table pid_ns_ctl_table[] = {
+       {
+               .procname = "ns_last_pid",
+               .maxlen = sizeof(int),
+               .mode = 0666, /* permissions are checked in the handler */
+               .proc_handler = pid_ns_ctl_handler,
+       },
+       { }
+};
+
+static struct ctl_path kern_path[] = { { .procname = "kernel", }, { } };
+
+int reboot_pid_ns(struct pid_namespace *pid_ns, int cmd)
+{
+       if (pid_ns == &init_pid_ns)
+               return 0;
+
+       switch (cmd) {
+       case LINUX_REBOOT_CMD_RESTART2:
+       case LINUX_REBOOT_CMD_RESTART:
+               pid_ns->reboot = SIGHUP;
+               break;
+
+       case LINUX_REBOOT_CMD_POWER_OFF:
+       case LINUX_REBOOT_CMD_HALT:
+               pid_ns->reboot = SIGINT;
+               break;
+       default:
+               return -EINVAL;
+       }
+
+       read_lock(&tasklist_lock);
+       force_sig(SIGKILL, pid_ns->child_reaper);
+       read_unlock(&tasklist_lock);
+
+       do_exit(0);
+
+       /* Not reached */
+       return 0;
+}
+
 static __init int pid_namespaces_init(void)
 {
        pid_ns_cachep = KMEM_CACHE(pid_namespace, SLAB_PANIC);
+       register_sysctl_paths(kern_path, pid_ns_ctl_table);
        return 0;
 }