All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Steven Rostedt <rostedt@goodmis.org>
Cc: linux-kernel@vger.kernel.org, Ingo Molnar <mingo@kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Frederic Weisbecker <fweisbec@gmail.com>,
	Jiri Olsa <jolsa@redhat.com>
Subject: Re: [RFC][PATCH 01/18 v2] ftrace: Add hash list to save RCU unsafe functions
Date: Tue, 3 Sep 2013 18:24:04 -0700	[thread overview]
Message-ID: <20130904012404.GI3871@linux.vnet.ibm.com> (raw)
In-Reply-To: <20130903195705.3ba3442f@gandalf.local.home>

On Tue, Sep 03, 2013 at 07:57:05PM -0400, Steven Rostedt wrote:
> On Tue, 3 Sep 2013 15:18:08 -0700
> "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> wrote:
> 
> 
> > > Just found this bug. Strange that gcc never gave me a warning :-/
> > 
> > I can't give gcc too much trouble, as I also didn't give you an
> > uninitialized-variable warning.
> 
> I was also chasing down a nasty bug that looked to be a pointer
> corruption somewhere. Still never found exactly where it happened, but
> it always happened with the following conditions:
> 
> synchronize_sched() was in progress
> 
> The ftrace_unsafe_callback() was preempted by an interrupt
> 
> lockdep was processing a lock
> 
> 
> A crash would happen which had memory corruption involved. But the
> above seemed always to be in play.
> 
> Now, I changed the logic from disabling context level recursion to
> disabling recursion at all. That is, the original code had used a
> series of bits to test for recursion (via helper functions) that would
> allow for the callback to be preempted by an interrupt, and then be
> called again.
> 
> The new code sets a per_cpu flag, and will not allow the callback to
> recurse if it were preempted by an interrupt. No real need to allow for
> that anyway.
> 
> I can go and debug this further, because I'm nervous that lockdep may
> have some kind of bug that the function tracer can trigger. But I'm not
> too concerned about it.
> 
> Here's the diff of the new code against the previous code.
> 
> Paul, can I still keep all your acks and reviewed-bys on this?

Yep, but some questions below.

							Thanx, Paul

> -- Steve
> 
> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
> index 310b727..899c8c1 100644
> --- a/kernel/trace/ftrace.c
> +++ b/kernel/trace/ftrace.c
> @@ -1373,7 +1373,7 @@ ftrace_hash_rec_enable(struct ftrace_ops *ops, int filter_hash);
> 
>  static int ftrace_convert_size_to_bits(int size)
>  {
> -	int bits;
> +	int bits = 0;
> 
>  	/*
>  	 * Make the hash size about 1/2 the # found
> diff --git a/kernel/trace/trace_functions.c b/kernel/trace/trace_functions.c
> index f5a9031..0883069 100644
> --- a/kernel/trace/trace_functions.c
> +++ b/kernel/trace/trace_functions.c
> @@ -561,16 +561,21 @@ static inline int init_func_cmd_traceon(void)
> 
>  #ifdef CONFIG_FTRACE_UNSAFE_RCU_CHECKER
> 
> +static DEFINE_PER_CPU(int, ftrace_rcu_running);
>  static atomic_t ftrace_unsafe_rcu_disabled;
> 
>  void ftrace_unsafe_rcu_checker_disable(void)
>  {
>  	atomic_inc(&ftrace_unsafe_rcu_disabled);
> +	/* Make sure the update is seen immediately */
> +	smp_wmb();

	smp_mb__after_atomic_inc()?

>  }
> 
>  void ftrace_unsafe_rcu_checker_enable(void)
>  {
>  	atomic_dec(&ftrace_unsafe_rcu_disabled);
> +	/* Make sure the update is seen immediately */
> +	smp_wmb();

	smp_mb__after_atomic_dec()?

>  }
> 
>  static DEFINE_PER_CPU(unsigned long, ftrace_rcu_func);
> @@ -588,15 +593,14 @@ static void
>  ftrace_unsafe_callback(unsigned long ip, unsigned long parent_ip,
>  		       struct ftrace_ops *op, struct pt_regs *pt_regs)
>  {
> -	int bit;
> -
> +	/* Make sure we see disabled or not first */
> +	smp_rmb();

	smp_mb__before_atomic_inc()?

>  	if (atomic_read(&ftrace_unsafe_rcu_disabled))
>  		return;
> 
>  	preempt_disable_notrace();
> 
> -	bit = trace_test_and_set_recursion(TRACE_FTRACE_START, TRACE_FTRACE_MAX);
> -	if (bit < 0)
> +	if (this_cpu_read(ftrace_rcu_running))
>  		goto out;
> 
>  	if (WARN_ONCE(ftrace_rcu_unsafe(ip),
> @@ -604,13 +608,15 @@ ftrace_unsafe_callback(unsigned long ip, unsigned long parent_ip,
>  		      (void *)ip))
>  		goto out;
> 
> +	this_cpu_write(ftrace_rcu_running, 1);
>  	this_cpu_write(ftrace_rcu_func, ip);
> +
>  	/* Should trigger a RCU splat if called from unsafe RCU function */
>  	rcu_read_lock();
>  	rcu_read_unlock();
>  	this_cpu_write(ftrace_rcu_func, 0);
> 
> -	trace_clear_recursion(bit);
> +	this_cpu_write(ftrace_rcu_running, 0);
>   out:
>  	preempt_enable_notrace();
>  }
> 


  parent reply	other threads:[~2013-09-04  1:24 UTC|newest]

Thread overview: 74+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-08-31  5:11 [RFC][PATCH 00/18 v2] ftrace/rcu: Handle unsafe RCU functions and ftrace callbacks Steven Rostedt
2013-08-31  5:11 ` [RFC][PATCH 01/18 v2] ftrace: Add hash list to save RCU unsafe functions Steven Rostedt
2013-08-31 19:28   ` Paul E. McKenney
2013-09-03 21:15   ` Steven Rostedt
2013-09-03 22:18     ` Paul E. McKenney
2013-09-03 23:57       ` Steven Rostedt
2013-09-04  0:18         ` Steven Rostedt
2013-09-04  1:11           ` [RFC][PATCH 19/18] ftrace: Print a message when the rcu checker is disabled Steven Rostedt
2013-09-04  1:25             ` Paul E. McKenney
2013-09-04  1:24         ` Paul E. McKenney [this message]
2013-09-04  1:51           ` [RFC][PATCH 01/18 v2] ftrace: Add hash list to save RCU unsafe functions Steven Rostedt
2013-09-04  1:56             ` Steven Rostedt
2013-09-04  2:01           ` Steven Rostedt
2013-09-04  2:03             ` Steven Rostedt
2013-09-04  4:18               ` Paul E. McKenney
2013-09-04 11:50                 ` Steven Rostedt
2013-09-04  4:12             ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 02/18 v2] ftrace: Do not set ftrace records for unsafe RCU when not allowed Steven Rostedt
2013-08-31 19:29   ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 03/18 v2] ftrace: Set ftrace internal function tracing RCU safe Steven Rostedt
2013-08-31 19:30   ` Paul E. McKenney
2013-08-31 19:44   ` Paul E. McKenney
2013-09-03 13:22     ` Steven Rostedt
2013-09-03 13:54       ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 04/18 v2] ftrace: Add test for ops against unsafe RCU functions in callback Steven Rostedt
2013-08-31 19:45   ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 05/18 v2] ftrace: Do not display non safe RCU functions in available_filter_functions Steven Rostedt
2013-08-31 19:46   ` Paul E. McKenney
2013-08-31 20:35     ` Steven Rostedt
2013-08-31 20:54       ` Paul E. McKenney
2013-09-03 13:26         ` Steven Rostedt
2013-09-03 13:54           ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 06/18 v2] ftrace: Add rcu_unsafe_filter_functions file Steven Rostedt
2013-08-31 19:46   ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 07/18 v2] ftrace: Add selftest to check if RCU unsafe functions are filtered properly Steven Rostedt
2013-08-31 19:47   ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 08/18 v2] ftrace/rcu: Do not trace debug_lockdep_rcu_enabled() Steven Rostedt
2013-08-31 19:21   ` Paul E. McKenney
2013-08-31 20:31     ` Steven Rostedt
2013-08-31  5:11 ` [RFC][PATCH 09/18 v2] ftrace: Fix a slight race in modifying what function callback gets traced Steven Rostedt
2013-08-31  5:11 ` [RFC][PATCH 10/18 v2] ftrace: Create a RCU unsafe checker Steven Rostedt
2013-08-31 19:49   ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 11/18 v2] ftrace: Adde infrastructure to stop RCU unsafe checker from checking Steven Rostedt
2013-08-31 19:52   ` Paul E. McKenney
2013-08-31 20:40     ` Steven Rostedt
2013-09-03 13:43     ` Steven Rostedt
2013-09-03 15:22       ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 12/18 v2] ftrace: Disable RCU unsafe checker when function graph is enabled Steven Rostedt
2013-08-31 19:55   ` Paul E. McKenney
2013-08-31 20:42     ` Steven Rostedt
2013-08-31 20:58       ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 13/18 v2] ftrace: Disable the RCU unsafe checker when irqsoff " Steven Rostedt
2013-08-31 19:58   ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 14/18 v2] ftrace/lockdep: Have the RCU lockdep splat show what function triggered Steven Rostedt
2013-08-31 19:59   ` Paul E. McKenney
2013-09-05 19:18   ` Peter Zijlstra
2013-09-05 19:52     ` Steven Rostedt
2013-09-06 12:57       ` Ingo Molnar
2013-09-06 13:16         ` Steven Rostedt
2013-09-05 19:35   ` Peter Zijlstra
2013-09-05 20:27     ` Steven Rostedt
2013-08-31  5:11 ` [RFC][PATCH 15/18 v2] ftrace/rcu: Mark functions that are RCU unsafe Steven Rostedt
2013-08-31 20:00   ` Paul E. McKenney
2013-08-31 20:43     ` Steven Rostedt
2013-08-31 20:54       ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 16/18 v2] rcu/irq/x86: " Steven Rostedt
2013-08-31 20:01   ` Paul E. McKenney
2013-08-31  5:11 ` [RFC][PATCH 17/18 v2] ftrace/cpuidle/x86: " Steven Rostedt
2013-08-31 11:07   ` Wysocki, Rafael J
2013-08-31 20:02   ` Paul E. McKenney
2013-09-04  0:16   ` H. Peter Anvin
2013-08-31  5:11 ` [RFC][PATCH 18/18 v2] ftrace/sched: " Steven Rostedt
2013-08-31 20:01   ` Paul E. McKenney
2013-08-31 15:50 ` [RFC][PATCH 00/18 v2] ftrace/rcu: Handle unsafe RCU functions and ftrace callbacks Steven Rostedt

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20130904012404.GI3871@linux.vnet.ibm.com \
    --to=paulmck@linux.vnet.ibm.com \
    --cc=akpm@linux-foundation.org \
    --cc=fweisbec@gmail.com \
    --cc=jolsa@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@kernel.org \
    --cc=peterz@infradead.org \
    --cc=rostedt@goodmis.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.