kernel/fork.c:copy_process(): don't add the uninitialized child to thread/task/pid lists

copy_process() adds the new child to thread_group/init_task.tasks list and
then does attach_pid(child, PIDTYPE_PID).  This means that the lockless
next_thread() or next_task() can see this thread with the wrong pid.  Say,
"ls /proc/pid/task" can list the same inode twice.

We could move attach_pid(child, PIDTYPE_PID) up, but in this case
find_task_by_vpid() can find the new thread before it was fully
initialized.

And this is already true for PIDTYPE_PGID/PIDTYPE_SID, With this patch
copy_process() initializes child->pids[*].pid first, then calls
attach_pid() to insert the task into the pid->tasks list.

attach_pid() no longer need the "struct pid*" argument, it is always
called after pid_link->pid was already set.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Sergey Dyasly <dserrg@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

2 files changed, 17 insertions, 11 deletions

diff --git a/kernel/fork.c b/kernel/fork.c
index 417cb864e20c..7d6962fb6156 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1121,6 +1121,12 @@ static void posix_cpu_timers_init(struct task_struct *tsk)
 	INIT_LIST_HEAD(&tsk->cpu_timers[2]);
 }
 
+static inline void
+init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
+{
+	 task->pids[type].pid = pid;
+}
+
 /*
  * This creates a new process as a copy of the old one,
  * but does not actually start it yet.
@@ -1449,7 +1455,11 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 	if (likely(p->pid)) {
 		ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
 
+		init_task_pid(p, PIDTYPE_PID, pid);
 		if (thread_group_leader(p)) {
+			init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
+			init_task_pid(p, PIDTYPE_SID, task_session(current));
+
 			if (is_child_reaper(pid)) {
 				ns_of_pid(pid)->child_reaper = p;
 				p->signal->flags |= SIGNAL_UNKILLABLE;
@@ -1457,10 +1467,10 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 
 			p->signal->leader_pid = pid;
 			p->signal->tty = tty_kref_get(current->signal->tty);
-			attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
-			attach_pid(p, PIDTYPE_SID, task_session(current));
 			list_add_tail(&p->sibling, &p->real_parent->children);
 			list_add_tail_rcu(&p->tasks, &init_task.tasks);
+			attach_pid(p, PIDTYPE_PGID);
+			attach_pid(p, PIDTYPE_SID);
 			__this_cpu_inc(process_counts);
 		} else {
 			current->signal->nr_threads++;
@@ -1470,7 +1480,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 			list_add_tail_rcu(&p->thread_group,
 					  &p->group_leader->thread_group);
 		}
-		attach_pid(p, PIDTYPE_PID, pid);
+		attach_pid(p, PIDTYPE_PID);
 		nr_threads++;
 	}
 
diff --git a/kernel/pid.c b/kernel/pid.c
index 0db3e791a06d..61980cefb1f5 100644
--- a/kernel/pid.c
+++ b/kernel/pid.c
@@ -373,14 +373,10 @@ EXPORT_SYMBOL_GPL(find_vpid);
 /*
  * attach_pid() must be called with the tasklist_lock write-held.
  */
-void attach_pid(struct task_struct *task, enum pid_type type,
-		struct pid *pid)
+void attach_pid(struct task_struct *task, enum pid_type type)
 {
-	struct pid_link *link;
-
-	link = &task->pids[type];
-	link->pid = pid;
-	hlist_add_head_rcu(&link->node, &pid->tasks[type]);
+	struct pid_link *link = &task->pids[type];
+	hlist_add_head_rcu(&link->node, &link->pid->tasks[type]);
 }
 
 static void __change_pid(struct task_struct *task, enum pid_type type,
@@ -412,7 +408,7 @@ void change_pid(struct task_struct *task, enum pid_type type,
 		struct pid *pid)
 {
 	__change_pid(task, type, pid);
-	attach_pid(task, type, pid);
+	attach_pid(task, type);
 }
 
 /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */