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;
Message-Id: <20200308222610.245444311@linutronix.de>
User-Agent: quilt/0.65
Date:   Sun, 08 Mar 2020 23:24:12 +0100
From:   Thomas Gleixner <tglx@linutronix.de>
To:     LKML <linux-kernel@vger.kernel.org>
Cc:     x86@kernel.org, Steven Rostedt <rostedt@goodmis.org>,
        Brian Gerst <brgerst@gmail.com>,
        Juergen Gross <jgross@suse.com>,
        Frederic Weisbecker <frederic@kernel.org>,
        Alexandre Chartre <alexandre.chartre@oracle.com>
Subject: [patch part-II V2 13/13] x86/entry/common: Split irq tracing in prepare_exit_to_usermode()
References: <20200308222359.370649591@linutronix.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

As in entry from user mode, lockdep and tracing have different
requirements. lockdep needs to know about the interrupts off state accross
the call to user_enter_irqsoff() but tracing is unsafe after the call.

Split it up and tell the tracer that interrupts are going to be enabled
before calling user_enter_irqsoff() and tell lockdep afterwards.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/entry/common.c |   12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -251,9 +251,19 @@ static noinline void __prepare_exit_to_u
 {
 	__prepare_exit_to_usermode(regs);
 
+	/*
+	 * Return to user space enables interrupts. Tell the tracer before
+	 * invoking user_enter_irqsoff() which switches to CONTEXT_USER and
+	 * RCU to rcuidle state. Lockdep still needs to keep the irqs
+	 * disabled state.
+	 */
+	__trace_hardirqs_on();
+
 	user_enter_irqoff();
 	mds_user_clear_cpu_buffers();
-	trace_hardirqs_on();
+
+	/* All done. Tell lockdep as well. */
+	lockdep_hardirqs_on(CALLER_ADDR0);
 }
 NOKPROBE_SYMBOL(prepare_exit_to_usermode);