Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Fri, 21 Jun 2019 19:19:20 +0200 (CEST)
From:   Thomas Gleixner <tglx@linutronix.de>
To:     Zhenzhong Duan <zhenzhong.duan@oracle.com>
cc:     linux-kernel@vger.kernel.org, mingo@kernel.org, bp@alien8.de,
        x86@kernel.org, hpa@zytor.com, jgross@suse.com,
        ndesaulniers@google.com, gregkh@linuxfoundation.org,
        srinivas.eeda@oracle.com
Subject: Re: [PATCH] x86/speculation/mds: Flush store buffer after wake up
 from sleep
In-Reply-To: <1561011237-12312-1-git-send-email-zhenzhong.duan@oracle.com>
Message-ID: <alpine.DEB.2.21.1906211916000.5503@nanos.tec.linutronix.de>
References: <1561011237-12312-1-git-send-email-zhenzhong.duan@oracle.com>
User-Agent: Alpine 2.21 (DEB 202 2017-01-01)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, 19 Jun 2019, Zhenzhong Duan wrote:

> Intel document says: "When a thread wakes from a sleep state, the store
> buffer is repartitioned again. This causes the store buffer to transfer
> store buffer entries from the thread that was already active to the one
> which just woke up."
> 
> To avoid data leak from sibling thread to the woken thread, flush store
> buffer right after wake up.

That's a pointless exercise. The buffers are flushed again when returning
to user space. Inside the kernel the potential leak is completely
uninteresting unless you consider the kernel as a malicious entity.
 
> Move mds_idle_clear_cpu_buffers() after trace_hardirqs_on() to ensure
> all store buffer entries are flushed before sleep.

I'm fine with that change, albeit trace_hardirqs_on() is hardly leaking
somethimg interesting.

Thanks,

	tglx