All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Paul E. McKenney" <paulmck@linux.ibm.com>
To: Scott Wood <swood@redhat.com>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
	Peter Zijlstra <peterz@infradead.org>,
	Steven Rostedt <rostedt@goodmis.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Juri Lelli <juri.lelli@redhat.com>,
	Clark Williams <williams@redhat.com>,
	linux-rt-users@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH RT 4/4] rcutorture: Avoid problematic critical section nesting
Date: Thu, 20 Jun 2019 14:18:26 -0700	[thread overview]
Message-ID: <20190620211826.GX26519@linux.ibm.com> (raw)
In-Reply-To: <20190619011908.25026-5-swood@redhat.com>

On Tue, Jun 18, 2019 at 08:19:08PM -0500, Scott Wood wrote:
> rcutorture was generating some nesting scenarios that are not
> reasonable.  Constrain the state selection to avoid them.
> 
> Example #1:
> 
> 1. preempt_disable()
> 2. local_bh_disable()
> 3. preempt_enable()
> 4. local_bh_enable()
> 
> On PREEMPT_RT, BH disabling takes a local lock only when called in
> non-atomic context.  Thus, atomic context must be retained until after BH
> is re-enabled.  Likewise, if BH is initially disabled in non-atomic
> context, it cannot be re-enabled in atomic context.
> 
> Example #2:
> 
> 1. rcu_read_lock()
> 2. local_irq_disable()
> 3. rcu_read_unlock()
> 4. local_irq_enable()
> 
> If the thread is preempted between steps 1 and 2,
> rcu_read_unlock_special.b.blocked will be set, but it won't be
> acted on in step 3 because IRQs are disabled.  Thus, reporting of the
> quiescent state will be delayed beyond the local_irq_enable().
> 
> Example #3:
> 
> 1. preempt_disable()
> 2. local_irq_disable()
> 3. preempt_enable()
> 4. local_irq_enable()
> 
> If need_resched is set between steps 1 and 2, then the reschedule
> in step 3 will not happen.
> 
> Signed-off-by: Scott Wood <swood@redhat.com>

OK for -rt, but as long as people can code those sequences without getting
their wrists slapped, RCU needs to deal with it.  So I cannot accept
this in mainline at the current time.  Yes, I will know when it is safe
to accept it when rcutorture's virtual wrist gets slapped in mainline.
Why did you ask?  ;-)

But I have to ask...  With this elaboration, is it time to make this a
data-driven state machine?  Or is the complexity not yet to the point
where that would constitute a simplification?

							Thanx, Paul

> ---
> TODO: Document restrictions and add debug checks for invalid sequences.
> 
> I had been planning to resolve #1 (only as shown, not the case of
> disabling preemption while non-atomic and enabling while atomic) by
> changing how migrate_disable() works to avoid the split behavior, but
> recently BH disabling was changed to do the same thing.  I still plan to
> send the migrate disable changes as a separate patchset, for the sake of
> the significant performance improvement I saw.
> ---
>  kernel/rcu/rcutorture.c | 92 +++++++++++++++++++++++++++++++++++++++++--------
>  1 file changed, 78 insertions(+), 14 deletions(-)
> 
> diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
> index 584b0d1da0a3..0523d9e78246 100644
> --- a/kernel/rcu/rcutorture.c
> +++ b/kernel/rcu/rcutorture.c
> @@ -73,10 +73,13 @@
>  #define RCUTORTURE_RDR_RBH	 0x08	/*  ... rcu_read_lock_bh(). */
>  #define RCUTORTURE_RDR_SCHED	 0x10	/*  ... rcu_read_lock_sched(). */
>  #define RCUTORTURE_RDR_RCU	 0x20	/*  ... entering another RCU reader. */
> -#define RCUTORTURE_RDR_NBITS	 6	/* Number of bits defined above. */
> +#define RCUTORTURE_RDR_ATOM_BH	 0x40	/*  ... disabling bh while atomic */
> +#define RCUTORTURE_RDR_ATOM_RBH	 0x80	/*  ... RBH while atomic */
> +#define RCUTORTURE_RDR_NBITS	 8	/* Number of bits defined above. */
>  #define RCUTORTURE_MAX_EXTEND	 \
>  	(RCUTORTURE_RDR_BH | RCUTORTURE_RDR_IRQ | RCUTORTURE_RDR_PREEMPT | \
> -	 RCUTORTURE_RDR_RBH | RCUTORTURE_RDR_SCHED)
> +	 RCUTORTURE_RDR_RBH | RCUTORTURE_RDR_SCHED | \
> +	 RCUTORTURE_RDR_ATOM_BH | RCUTORTURE_RDR_ATOM_RBH)
>  #define RCUTORTURE_RDR_MAX_LOOPS 0x7	/* Maximum reader extensions. */
>  					/* Must be power of two minus one. */
>  #define RCUTORTURE_RDR_MAX_SEGS (RCUTORTURE_RDR_MAX_LOOPS + 3)
> @@ -1111,31 +1114,52 @@ static void rcutorture_one_extend(int *readstate, int newstate,
>  	WARN_ON_ONCE((idxold >> RCUTORTURE_RDR_SHIFT) > 1);
>  	rtrsp->rt_readstate = newstate;
>  
> -	/* First, put new protection in place to avoid critical-section gap. */
> +	/*
> +	 * First, put new protection in place to avoid critical-section gap.
> +	 * Disable preemption around the ATOM disables to ensure that
> +	 * in_atomic() is true.
> +	 */
>  	if (statesnew & RCUTORTURE_RDR_BH)
>  		local_bh_disable();
> +	if (statesnew & RCUTORTURE_RDR_RBH)
> +		rcu_read_lock_bh();
>  	if (statesnew & RCUTORTURE_RDR_IRQ)
>  		local_irq_disable();
>  	if (statesnew & RCUTORTURE_RDR_PREEMPT)
>  		preempt_disable();
> -	if (statesnew & RCUTORTURE_RDR_RBH)
> -		rcu_read_lock_bh();
>  	if (statesnew & RCUTORTURE_RDR_SCHED)
>  		rcu_read_lock_sched();
> +	preempt_disable();
> +	if (statesnew & RCUTORTURE_RDR_ATOM_BH)
> +		local_bh_disable();
> +	if (statesnew & RCUTORTURE_RDR_ATOM_RBH)
> +		rcu_read_lock_bh();
> +	preempt_enable();
>  	if (statesnew & RCUTORTURE_RDR_RCU)
>  		idxnew = cur_ops->readlock() << RCUTORTURE_RDR_SHIFT;
>  
> -	/* Next, remove old protection, irq first due to bh conflict. */
> +	/*
> +	 * Next, remove old protection, in decreasing order of strength
> +	 * to avoid unlock paths that aren't safe in the stronger
> +	 * context.  Disable preemption around the ATOM enables in
> +	 * case the context was only atomic due to IRQ disabling.
> +	 */
> +	preempt_disable();
>  	if (statesold & RCUTORTURE_RDR_IRQ)
>  		local_irq_enable();
> -	if (statesold & RCUTORTURE_RDR_BH)
> +	if (statesold & RCUTORTURE_RDR_ATOM_BH)
>  		local_bh_enable();
> +	if (statesold & RCUTORTURE_RDR_ATOM_RBH)
> +		rcu_read_unlock_bh();
> +	preempt_enable();
>  	if (statesold & RCUTORTURE_RDR_PREEMPT)
>  		preempt_enable();
> -	if (statesold & RCUTORTURE_RDR_RBH)
> -		rcu_read_unlock_bh();
>  	if (statesold & RCUTORTURE_RDR_SCHED)
>  		rcu_read_unlock_sched();
> +	if (statesold & RCUTORTURE_RDR_BH)
> +		local_bh_enable();
> +	if (statesold & RCUTORTURE_RDR_RBH)
> +		rcu_read_unlock_bh();
>  	if (statesold & RCUTORTURE_RDR_RCU)
>  		cur_ops->readunlock(idxold >> RCUTORTURE_RDR_SHIFT);
>  
> @@ -1171,6 +1195,12 @@ static int rcutorture_extend_mask_max(void)
>  	int mask = rcutorture_extend_mask_max();
>  	unsigned long randmask1 = torture_random(trsp) >> 8;
>  	unsigned long randmask2 = randmask1 >> 3;
> +	unsigned long preempts = RCUTORTURE_RDR_PREEMPT | RCUTORTURE_RDR_SCHED;
> +	unsigned long preempts_irq = preempts | RCUTORTURE_RDR_IRQ;
> +	unsigned long nonatomic_bhs = RCUTORTURE_RDR_BH | RCUTORTURE_RDR_RBH;
> +	unsigned long atomic_bhs = RCUTORTURE_RDR_ATOM_BH |
> +				   RCUTORTURE_RDR_ATOM_RBH;
> +	unsigned long tmp;
>  
>  	WARN_ON_ONCE(mask >> RCUTORTURE_RDR_SHIFT);
>  	/* Most of the time lots of bits, half the time only one bit. */
> @@ -1178,11 +1208,45 @@ static int rcutorture_extend_mask_max(void)
>  		mask = mask & randmask2;
>  	else
>  		mask = mask & (1 << (randmask2 % RCUTORTURE_RDR_NBITS));
> -	/* Can't enable bh w/irq disabled. */
> -	if ((mask & RCUTORTURE_RDR_IRQ) &&
> -	    ((!(mask & RCUTORTURE_RDR_BH) && (oldmask & RCUTORTURE_RDR_BH)) ||
> -	     (!(mask & RCUTORTURE_RDR_RBH) && (oldmask & RCUTORTURE_RDR_RBH))))
> -		mask |= RCUTORTURE_RDR_BH | RCUTORTURE_RDR_RBH;
> +
> +	/*
> +	 * Can't enable bh w/irq disabled.
> +	 *
> +	 * Can't enable preemption with irqs disabled, if irqs had ever
> +	 * been enabled during this preempt critical section (could miss
> +	 * a reschedule).
> +	 */
> +	tmp = atomic_bhs | nonatomic_bhs | preempts;
> +	if (mask & RCUTORTURE_RDR_IRQ)
> +		mask |= oldmask & tmp;
> +
> +	/*
> +	 * Can't release the outermost rcu lock in an irq disabled
> +	 * section without preemption also being disabled, if irqs had
> +	 * ever been enabled during this RCU critical section (could leak
> +	 * a special flag and delay reporting the qs).
> +	 */
> +	if ((oldmask & RCUTORTURE_RDR_RCU) && (mask & RCUTORTURE_RDR_IRQ) &&
> +	    !(mask & preempts))
> +		mask |= RCUTORTURE_RDR_RCU;
> +
> +	/* Can't modify atomic bh in non-atomic context */
> +	if ((oldmask & atomic_bhs) && (mask & atomic_bhs) &&
> +	    !(mask & preempts_irq)) {
> +		mask |= oldmask & preempts_irq;
> +		if (mask & RCUTORTURE_RDR_IRQ)
> +			mask |= oldmask & tmp;
> +	}
> +	if ((mask & atomic_bhs) && !(mask & preempts_irq))
> +		mask |= RCUTORTURE_RDR_PREEMPT;
> +
> +	/* Can't modify non-atomic bh in atomic context */
> +	tmp = nonatomic_bhs;
> +	if (oldmask & preempts_irq)
> +		mask &= ~tmp;
> +	if ((oldmask | mask) & preempts_irq)
> +		mask |= oldmask & tmp;
> +
>  	if ((mask & RCUTORTURE_RDR_IRQ) &&
>  	    !(mask & cur_ops->ext_irq_conflict) &&
>  	    (oldmask & cur_ops->ext_irq_conflict))
> -- 
> 1.8.3.1
> 

  reply	other threads:[~2019-06-20 21:19 UTC|newest]

Thread overview: 31+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-06-19  1:19 [PATCH RT 0/4] Address rcutorture issues Scott Wood
2019-06-19  1:19 ` [PATCH RT 1/4] rcu: Acquire RCU lock when disabling BHs Scott Wood
2019-06-20 20:53   ` Paul E. McKenney
2019-06-20 21:06     ` Scott Wood
2019-06-20 21:20       ` Paul E. McKenney
2019-06-20 21:38         ` Scott Wood
2019-06-20 22:16           ` Paul E. McKenney
2019-06-19  1:19 ` [PATCH RT 2/4] sched: migrate_enable: Use sleeping_lock to indicate involuntary sleep Scott Wood
2019-06-19  1:19 ` [RFC PATCH RT 3/4] rcu: unlock special: Treat irq and preempt disabled the same Scott Wood
2019-06-20 21:10   ` Paul E. McKenney
2019-06-20 21:59     ` Scott Wood
2019-06-20 22:25       ` Paul E. McKenney
2019-06-20 23:08         ` Scott Wood
2019-06-22  0:26           ` Paul E. McKenney
2019-06-22 19:13             ` Paul E. McKenney
2019-06-24 17:40               ` Scott Wood
2019-06-19  1:19 ` [RFC PATCH RT 4/4] rcutorture: Avoid problematic critical section nesting Scott Wood
2019-06-20 21:18   ` Paul E. McKenney [this message]
2019-06-20 21:43     ` Scott Wood
2019-06-21 16:38     ` Sebastian Andrzej Siewior
2019-06-21 23:59       ` Paul E. McKenney
2019-06-26 15:08         ` Steven Rostedt
2019-06-26 16:49           ` Scott Wood
2019-06-27 18:00             ` Paul E. McKenney
2019-06-27 20:16               ` Scott Wood
2019-06-27 20:50                 ` Paul E. McKenney
2019-06-27 22:46                   ` Scott Wood
2019-06-28  0:52                     ` Paul E. McKenney
2019-06-28 19:37                       ` Scott Wood
2019-06-28 20:24                         ` Paul E. McKenney
2019-06-20 19:12 ` [PATCH RT 0/4] Address rcutorture issues Paul E. McKenney

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=20190620211826.GX26519@linux.ibm.com \
    --to=paulmck@linux.ibm.com \
    --cc=bigeasy@linutronix.de \
    --cc=juri.lelli@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-rt-users@vger.kernel.org \
    --cc=peterz@infradead.org \
    --cc=rostedt@goodmis.org \
    --cc=swood@redhat.com \
    --cc=tglx@linutronix.de \
    --cc=williams@redhat.com \
    /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.