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;
MIME-Version: 1.0
References: <20190729010734.3352-1-devel@etsukata.com> <20190729112126.6554b141@gandalf.local.home>
 <2ceec933-503e-5d58-60b4-85b491b017d4@etsukata.com> <20190729221506.1aed7dfc@oasis.local.home>
In-Reply-To: <20190729221506.1aed7dfc@oasis.local.home>
From:   Joel Fernandes <joel@joelfernandes.org>
Date:   Tue, 30 Jul 2019 00:19:36 -0400
Message-ID: <CAEXW_YT_vh1iGOwXdymzPO4-k59=+tqQCWHHokv+Kfif5GHypw@mail.gmail.com>
Subject: Re: [PATCH] tracing: Prevent RCU EQS breakage in preemptirq events
To:     Steven Rostedt <rostedt@goodmis.org>
Cc:     Eiichi Tsukata <devel@etsukata.com>,
        "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
        Thomas Glexiner <tglx@linutronix.de>,
        Peter Zilstra <peterz@infradead.org>,
        Ingo Molnar <mingo@redhat.com>,
        Frederic Weisbecker <fweisbec@gmail.com>,
        Luto <luto@amacapital.net>, LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Mon, Jul 29, 2019 at 10:15 PM Steven Rostedt <rostedt@goodmis.org> wrote:
[snip]
> > If the problem was only with userstacktrace, it will be reasonable to
> > surround only the userstack unwinder. But the situation is similar to
> > the previous "tracing vs CR2" case. As Peter taught me in
> > https://lore.kernel.org/lkml/20190708074823.GV3402@hirez.programming.kicks-ass.net/
> > there are some other codes likely to to user access.
> > So I surround preemptirq events earlier.
>
> I disagree. The issue is with the attached callbacks that call
> something (a stack unwinder) that can fault.
>
> This is called whenever irqs are disabled. I say we surround the
> culprit (the stack unwinder or stack trace) and not punish the irq
> enable/disable events.

I agree with everything Steve said.

thanks,

 - Joel