pid: perform free_pid() calls outside of tasklist_lock

As the clone side already executes pid allocation with only pidmap_lock
held, issuing free_pid() while still holding tasklist_lock exacerbates
total hold time of the latter.

More things may show up later which require initial clean up with the
lock held and allow finishing without it. For that reason a struct to
collect such work is added instead of merely passing the pid array.

Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Mateusz Guzik <mjguzik@gmail.com>
Link: https://lore.kernel.org/r/20250206164415.450051-5-mjguzik@gmail.com
Acked-by: "Liam R. Howlett" <Liam.Howlett@Oracle.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>

1 files changed, 22 insertions, 22 deletions

diff --git a/kernel/pid.c b/kernel/pid.c
index 2ae872f689a7..73625f28c166 100644
--- a/kernel/pid.c
+++ b/kernel/pid.c
@@ -88,20 +88,6 @@ struct pid_namespace init_pid_ns = {
 };
 EXPORT_SYMBOL_GPL(init_pid_ns);
 
-/*
- * Note: disable interrupts while the pidmap_lock is held as an
- * interrupt might come in and do read_lock(&tasklist_lock).
- *
- * If we don't disable interrupts there is a nasty deadlock between
- * detach_pid()->free_pid() and another cpu that does
- * spin_lock(&pidmap_lock) followed by an interrupt routine that does
- * read_lock(&tasklist_lock);
- *
- * After we clean up the tasklist_lock and know there are no
- * irq handlers that take it we can leave the interrupts enabled.
- * For now it is easier to be safe than to prove it can't happen.
- */
-
 static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
 seqcount_spinlock_t pidmap_lock_seq = SEQCNT_SPINLOCK_ZERO(pidmap_lock_seq, &pidmap_lock);
 
@@ -128,10 +114,11 @@ static void delayed_put_pid(struct rcu_head *rhp)
 
 void free_pid(struct pid *pid)
 {
-	/* We can be called with write_lock_irq(&tasklist_lock) held */
 	int i;
 	unsigned long flags;
 
+	lockdep_assert_not_held(&tasklist_lock);
+
 	spin_lock_irqsave(&pidmap_lock, flags);
 	for (i = 0; i <= pid->level; i++) {
 		struct upid *upid = pid->numbers + i;
@@ -160,6 +147,18 @@ void free_pid(struct pid *pid)
 	call_rcu(&pid->rcu, delayed_put_pid);
 }
 
+void free_pids(struct pid **pids)
+{
+	int tmp;
+
+	/*
+	 * This can batch pidmap_lock.
+	 */
+	for (tmp = PIDTYPE_MAX; --tmp >= 0; )
+		if (pids[tmp])
+			free_pid(pids[tmp]);
+}
+
 struct pid *alloc_pid(struct pid_namespace *ns, pid_t *set_tid,
 		      size_t set_tid_size)
 {
@@ -347,8 +346,8 @@ void attach_pid(struct task_struct *task, enum pid_type type)
 	hlist_add_head_rcu(&task->pid_links[type], &pid->tasks[type]);
 }
 
-static void __change_pid(struct task_struct *task, enum pid_type type,
-			struct pid *new)
+static void __change_pid(struct pid **pids, struct task_struct *task,
+			 enum pid_type type, struct pid *new)
 {
 	struct pid **pid_ptr, *pid;
 	int tmp;
@@ -370,18 +369,19 @@ static void __change_pid(struct task_struct *task, enum pid_type type,
 		if (pid_has_task(pid, tmp))
 			return;
 
-	free_pid(pid);
+	WARN_ON(pids[type]);
+	pids[type] = pid;
 }
 
-void detach_pid(struct task_struct *task, enum pid_type type)
+void detach_pid(struct pid **pids, struct task_struct *task, enum pid_type type)
 {
-	__change_pid(task, type, NULL);
+	__change_pid(pids, task, type, NULL);
 }
 
-void change_pid(struct task_struct *task, enum pid_type type,
+void change_pid(struct pid **pids, struct task_struct *task, enum pid_type type,
 		struct pid *pid)
 {
-	__change_pid(task, type, pid);
+	__change_pid(pids, task, type, pid);
 	attach_pid(task, type);
 }