On Wed, Apr 29, 2020 at 04:22:35PM +0300, Dan Carpenter wrote:
> The rcu_torture_read_exit_init() function is supposed to return negative
> error codes which get propagated back down the call tree but the current
> code returns true on failure.
>
> Fixes: e02882cd57e3 ("rcutorture: Add races with task-exit processing")
> Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
Good catch! Talk about code sort of working by accident!!!
Thank you, and I folded this into the original commit with attribution.
Thanx, Paul
> ---
> kernel/rcu/rcutorture.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
> index 269881e51dc6d..5270674128029 100644
> --- a/kernel/rcu/rcutorture.c
> +++ b/kernel/rcu/rcutorture.c
> @@ -2434,10 +2434,10 @@ static int rcu_torture_read_exit(void *unused)
> return 0;
> }
>
> -static bool rcu_torture_read_exit_init(void)
> +static int rcu_torture_read_exit_init(void)
> {
> if (read_exit <= 0)
> - return true;
> + return -EINVAL;
> init_waitqueue_head(&read_exit_wq);
> read_exit_child_stop = false;
> read_exit_child_stopped = false;
> --
> 2.26.2
>
