Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Tue, 12 May 2020 20:16:14 +0200
From:   Oleg Nesterov <oleg@redhat.com>
To:     Davidlohr Bueso <dave@stgolabs.net>
Cc:     akpm@linux-foundation.org, peterz@infradead.org,
        paulmck@kernel.org, tglx@linutronix.de,
        linux-kernel@vger.kernel.org, Davidlohr Bueso <dbueso@suse.de>
Subject: Re: [PATCH 1/2] kernel/sys: only rely on rcu for getpriority(2)
Message-ID: <20200512181613.GD28621@redhat.com>
References: <20200512000353.23653-1-dave@stgolabs.net>
 <20200512000353.23653-2-dave@stgolabs.net>
 <20200512150936.GA28621@redhat.com>
 <20200512160915.n3plwrwwrlpfqyrs@linux-p48b>
 <20200512164130.GC28621@redhat.com>
 <20200512165824.t6ktwllqlvkiingv@linux-p48b>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200512165824.t6ktwllqlvkiingv@linux-p48b>
User-Agent: Mutt/1.5.24 (2015-08-30)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 05/12, Davidlohr Bueso wrote:
>
> Right, but setting the flag is an indication that the tasklist_lock
> will be taken

Yes,

> and removed from the list,

Well no. If this task is not a group leader, or if it is traced this won't
happen "soon", this will happen when parent or debugger call wait().

But this doesn't matter. Lets suppose that the task is always removed from
the list right after it sets PF_EXITING. Now,

> and therefore we could
> optimistically avoid considering that task altogether

Why?? This is what I can't understand.

If sys_getpriority() sees PF_EXITING, we can pretend it was called before
this task has exited, or even right before this flag was set. Why should we
skip this task?

And otoh, this check can not help in any case, PF_EXITING can be set right
after the check.

So I still think this check can only add the unnecessary confusion, even if
we forget about change in behaviour.

And finally, whatever I missed, I do not understand how this connects to
"avoid the tasklist_lock". Whether we want it or not does not depend on
the locking, at all.

Oleg.