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

[Oleg Nesterov <oleg@redhat.com>]

Hi Mingyu,

On 10/12, Mingyu Wang wrote:
>
> In the original report, the task triggering the warning was in process
> context, through a TTY ioctl. Here are the relevant excerpts; omissions
> are marked with [...].

Thanks... so according to lockdep we have the

	&dev->event_lock --> &f_owner->lock --> tasklist_lock

chain. At first glance this suggests that kill_fasync()->send_sigio() can
be called under input_dev->event_lock. But I know nothing about drivers/input,
so I can't find who does this...

And,

> input_inject_event+0xa4/0x3b0 drivers/input/input.c:419
...
> kbd_bh+0x23c/0x310 drivers/tty/vt/keyboard.c:1244
> tasklet_action_common+0x25c/0x3d0 kernel/softirq.c:925

means that ->event_lock can be acquired in softirq context.

That is all I can understand right now. Can you can send the full full
report without omissions? Not that I think this will help, at least to me ;)

Perhaps something like:

read_lock(tasklist)	lock(event_lock)
						write_lock(tasklist)
<Interrupt>
lock(event_lock)	// send_sigio()
			// ...
			read_lock(tasklist)
			
Oleg.