From: "Paul E. McKenney" <paulmck@xxxxxxxxxx>
The rcu_tasks_trace_postgp() function uses for_each_process_thread()
to scan the task list without the benefit of RCU read-side protection,
which can result in use-after-free errors on task_struct structures.
This error was missed because the TRACE01 rcutorture scenario enables
lockdep, but also builds with CONFIG_PREEMPT_NONE=y. In this situation,
preemption is disabled everywhere, so lockdep thinks everywhere can
be a legitimate RCU reader. This commit therefore adds the needed
rcu_read_lock() and rcu_read_unlock().
Note that this bug can occur only after an RCU Tasks Trace CPU stall
warning, which by default only happens after a grace period has extended
for ten minutes (yes, not a typo, minutes).
Fixes: 4593e772b502 ("rcu-tasks: Add stall warnings for RCU Tasks Trace")
Cc: Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx>
Cc: Daniel Borkmann <daniel@xxxxxxxxxxxxx>
Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
Cc: <bpf@xxxxxxxxxxxxxxx>
Cc: <stable@xxxxxxxxxxxxxxx> # 5.7.x
Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxx>
---
kernel/rcu/tasks.h | 2 ++
1 file changed, 2 insertions(+)
diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
index fcd9c25..d5d9f2d 100644
--- a/kernel/rcu/tasks.h
+++ b/kernel/rcu/tasks.h
@@ -1088,9 +1088,11 @@ static void rcu_tasks_trace_postgp(struct rcu_tasks *rtp)
if (ret)
break; // Count reached zero.
// Stall warning time, so make a list of the offenders.
+ rcu_read_lock();
for_each_process_thread(g, t)
if (READ_ONCE(t->trc_reader_special.b.need_qs))
trc_add_holdout(t, &holdouts);
+ rcu_read_unlock();
firstreport = true;
list_for_each_entry_safe(t, g, &holdouts, trc_holdout_list) {
if (READ_ONCE(t->trc_reader_special.b.need_qs))
--
2.9.5
