Re: [PATCH v2] fs/fcntl: fix SOFTIRQ-unsafe lock order in fasync signaling

[Oleg Nesterov]

On 10/12, Jeff Layton wrote:
>
> We probably should
> revert this patch for now, until the problem and potential solution are
> more clear.

Agreed... I'll send the revert for now.

Hmm. The "if (type <= PIDTYPE_TGID)" case in send_sigio/send_sigurg looks
wrong too. send_sigio_to_task() -> ... -> lock_task_sighand() can fail if
we race with de_thread().

Oleg.