[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20170104161712.GL3742@linux.vnet.ibm.com>
Date: Wed, 4 Jan 2017 08:17:12 -0800
From: "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To: Steven Rostedt <rostedt@...dmis.org>
Cc: Enrico Mioso <mrkiko.rs@...il.com>, linux-kernel@...r.kernel.org,
Josh Triplett <josh@...htriplett.org>,
Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
Subject: Re: INFO: rcu_sched self-detected stall on CPU
On Wed, Jan 04, 2017 at 09:16:31AM -0500, Steven Rostedt wrote:
> On Wed, 4 Jan 2017 05:46:08 -0800
> "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com> wrote:
>
> > I suggest enabling tracing for timers with a goal of working out why
> > the rcu_sched task is not being awakened regularly -- during a grace
> > period, it should be awakened every three jiffies or so (depending on
> > the value of HZ).
>
> Perhaps enable events for rcu, sched and irqs as well?
Good point -- but you will need to build with CONFIG_RCU_TRACE=y
to get most of the RCU events.
Thanx, Paul
Powered by blists - more mailing lists