Received-SPF: pass (google.com: domain of linux-kernel+bounces-86706-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Message-ID: <tencent_BE8C3169030CD9C2FC7548832C2994921609@qq.com>
From: linke <lilinke99@qq.com>
Content-Type: text/plain;
	charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3774.100.2.1.4\))
Subject: Re: [PATCH] ring-buffer: use READ_ONCE() to read
 cpu_buffer->commit_page in concurrent environment
Date: Thu, 29 Feb 2024 20:32:26 +0800
Cc: lilinke99@qq.com,
 linux-kernel@vger.kernel.org,
 linux-trace-kernel@vger.kernel.org,
 mathieu.desnoyers@efficios.com,
 mhiramat@kernel.org
To: rostedt@goodmis.org

Hi Steven, sorry for the late reply.

> 
> Now the reason for the above READ_ONCE() is because the variables *are*
> going to be used again. We do *not* want the compiler to play any games
> with that.
> 

I don't think it is because the variables are going to be used again. 
Compiler optimizations barely do bad things in single thread programs. It
is because cpu_buffer->commit_page may change concurrently and should be
accessed atomically.

	/* Make sure commit page didn't change */
	curr_commit_page = READ_ONCE(cpu_buffer->commit_page);
	curr_commit_ts = READ_ONCE(curr_commit_page->page->time_stamp);

	/* If the commit page changed, then there's more data */
	if (curr_commit_page != commit_page ||
	    curr_commit_ts != commit_ts)
		return 0;

This code read cpu_buffer->commit_page and time_stamp again to check
whether commit page changed. It shows that cpu_buffer->commit_page and 
time_stamp may be changed by other threads.

        commit_page = cpu_buffer->commit_page;
        commit_ts = commit_page->page->time_stamp;

So the commit_page and time_stamp above is read while other threads may
change it. I think it is a data race if it is not atomic. Thus it is 
necessary to use READ_ONCE() here.