2019-07-24 17:08:11

by Joel Fernandes

Subject: [PATCH] pidfd: Add warning if exit_state is 0 during notification

Previously a condition got missed where the pidfd waiters are awakened
before the exit_state gets set. This can result in a missed notification
[1] and the polling thread waiting forever.

It is fixed now, however it would be nice to avoid this kind of issue
going unnoticed in the future. So just add a warning to catch it in the
future.

[1] https://lore.kernel.org/lkml/[email protected]/

Signed-off-by: Joel Fernandes (Google) <[email protected]>
---
kernel/signal.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/kernel/signal.c b/kernel/signal.c
index 91b789dd6e72..349f5a67f100 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1885,6 +1885,7 @@ static void do_notify_pidfd(struct task_struct *task)
{
struct pid *pid;

+ WARN_ON(task->exit_state == 0);
pid = task_pid(task);
wake_up_all(&pid->wait_pidfd);
}
--
2.22.0.657.g960e92d24f-goog

2019-07-24 17:10:03

by Christian Brauner

Subject: Re: [PATCH] pidfd: Add warning if exit_state is 0 during notification

On July 24, 2019 6:48:16 PM GMT+02:00, "Joel Fernandes (Google)" <[email protected]> wrote:
>Previously a condition got missed where the pidfd waiters are awakened
>before the exit_state gets set. This can result in a missed
>notification
>[1] and the polling thread waiting forever.
>
>It is fixed now, however it would be nice to avoid this kind of issue
>going unnoticed in the future. So just add a warning to catch it in the
>future.
>
>[1]
>https://lore.kernel.org/lkml/[email protected]/
>
>Signed-off-by: Joel Fernandes (Google) <[email protected]>
>---
> kernel/signal.c | 1 +
> 1 file changed, 1 insertion(+)
>
>diff --git a/kernel/signal.c b/kernel/signal.c
>index 91b789dd6e72..349f5a67f100 100644
>--- a/kernel/signal.c
>+++ b/kernel/signal.c
>@@ -1885,6 +1885,7 @@ static void do_notify_pidfd(struct task_struct
>*task)
> {
> struct pid *pid;
>
>+ WARN_ON(task->exit_state == 0);
> pid = task_pid(task);
> wake_up_all(&pid->wait_pidfd);
> }

Makes sense to me.
I'm picking this up unless someone sees a problem with this patch.

Reviewed-by: Christian Brauner <[email protected]>