Re: [PATCH] mm,hwpoison: non-current task should be checked early_kill for force_early

[Oscar Salvador <osalvador@suse.de>]

On Fri, Jan 15, 2021 at 03:55:06PM +0800, Aili Yao wrote:
> Hello,From 740148005f1333150ee1e119ed8a34454abb7c1d Mon Sep 17 00:00:00 2001
> From: Aili Yao <yaoaili@kingsoft.com>
> Date: Fri, 15 Jan 2021 14:56:06 +0800
> Subject: [PATCH] mm,hwpoison: non-current task should be checked early_kill
>  for force_early
> 
> Other process may also care the error info with the early-kill flag set.
> when force_early is set, and if tsk is not current, leave it to the
> early-kill flag check.

I am having a hard time trying to grasp what are you trying to achieve here.
Could you elaborate some more? Ideally stating what is the problem you are
fixing here.

I was also a bit confused about task_early_kill.
So, if force_early is true (aka. MF_ACTION_REQUIRED was set), we only
return a task struct if we find the task or a thread belonging to the task
(sharing the mm).

If it is not set by the caller of memory_failure, we still want to check the
task/threads' MCE policy to check whether PF_MCE_KILL_EARLY it was set with
prctl, right?

That makes sense to me.

So, back to your change, if force_early was set, but the main task does not
match, you still want to check whether any thread belonging to the main task
has the policy set?

-- 
Oscar Salvador
SUSE L3