Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 31 May 2022 14:05:19 +0100
From:   Matthew Wilcox <willy@infradead.org>
To:     Miaohe Lin <linmiaohe@huawei.com>
Cc:     akpm@linux-foundation.org, naoya.horiguchi@nec.com,
        peterx@redhat.com, apopple@nvidia.com, ying.huang@intel.com,
        osalvador@suse.de, mike.kravetz@oracle.com,
        songmuchun@bytedance.com, hch@lst.de, dhowells@redhat.com,
        cl@linux.com, linux-mm@kvack.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4 1/4] mm: reduce the rcu lock duration
Message-ID: <YpYSjw/T8jTGr7e8@casper.infradead.org>
References: <20220530113016.16663-1-linmiaohe@huawei.com>
 <20220530113016.16663-2-linmiaohe@huawei.com>
 <YpYQ937N8Daad6u/@casper.infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YpYQ937N8Daad6u/@casper.infradead.org>
Precedence: bulk

On Tue, May 31, 2022 at 01:58:31PM +0100, Matthew Wilcox wrote:
> On Mon, May 30, 2022 at 07:30:13PM +0800, Miaohe Lin wrote:
> > Commit 3268c63eded4 ("mm: fix move/migrate_pages() race on task struct")
> > extends the period of the rcu_read_lock until after the permissions checks
> > are done to prevent the task pointed to from changing from under us. But
> > the task_struct refcount is also taken at that time, the reference to task
> > is guaranteed to be stable. So it's unnecessary to extend the period of
> > the rcu_read_lock. Release the rcu lock after task refcount is successfully
> > grabbed to reduce the rcu holding time.
> 
> But why bother?  You know the RCU read lock isn't a "real" lock, right?

Looking over this code some more, I think this may harm performance.
ptrace_may_access() itself takes the rcu_read_lock().  So we currently
have:

rcu_read_lock()
rcu_read_lock();
rcu_read_unlock();
rcu_read_unlock();

In at least one RCU configuration, rcu_read_lock() maps to
preempt_disable().  Nested preempt_disable() just bump a counter, while
that counter reaching zero incurs some actual work.  So nested
rcu_read_lock() can be better than sequential lock/unlock/lock/unlock.

This needs far better justification.