Re: [PATCH tip/core/rcu 01/45] rcutorture: Add forward-progress checking for writer

[josh@joshtriplett.org]

On Mon, Apr 28, 2014 at 05:24:49PM -0700, Paul E. McKenney wrote:
> From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
> 
> The rcutorture output currently does not distinguish between stalls in
> the RCU implementation and stalls in the rcu_torture_writer() kthreads.
> This commit therefore adds some diagnostics to help distinguish between
> these two conditions, at least for the non-SRCU implementations.  (SRCU
> does not provide evidence of update-side forward progress by design.)
> 
> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

The concept makes sense, and the writer state annotations seem like a
useful debugging mechanism, but having RCU know about RCU torture types
seems fundamentally wrong.  This mechanism accesses rcu_state, which is
already implementation-specific, so why not just only define the
function for the RCU implementations that support it, and then have a
function pointer in the torture-test structure to report a stall?

- Josh Triplett

>  include/linux/rcupdate.h | 19 +++++++++++++++++++
>  kernel/rcu/rcutorture.c  | 37 +++++++++++++++++++++++++++++++++++++
>  kernel/rcu/tree.c        | 18 ++++++++++++++++++
>  3 files changed, 74 insertions(+)
> 
> diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> index 00a7fd61b3c6..a6c3898e141e 100644
> --- a/include/linux/rcupdate.h
> +++ b/include/linux/rcupdate.h
> @@ -51,7 +51,17 @@ extern int rcu_expedited; /* for sysctl */
>  extern int rcutorture_runnable; /* for sysctl */
>  #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
>  
> +enum rcutorture_type {
> +	RTORT_BUSTED,
> +	RTORT_RCU,
> +	RTORT_RCU_BH,
> +	RTORT_RCU_SCHED,
> +	RTORT_SRCU
> +};
> +
>  #if defined(CONFIG_TREE_RCU) || defined(CONFIG_TREE_PREEMPT_RCU)
> +void rcutorture_get_gp_data(enum rcutorture_type test_type, int *flags,
> +			    unsigned long *gpnum, unsigned long *completed);
>  void rcutorture_record_test_transition(void);
>  void rcutorture_record_progress(unsigned long vernum);
>  void do_trace_rcu_torture_read(const char *rcutorturename,
> @@ -60,6 +70,15 @@ void do_trace_rcu_torture_read(const char *rcutorturename,
>  			       unsigned long c_old,
>  			       unsigned long c);
>  #else
> +static inline void rcutorture_get_gp_data(enum rcutorture_type test_type,
> +					  int *flags,
> +					  unsigned long *gpnum,
> +					  unsigned long *completed)
> +{
> +	*flags = 0;
> +	*gpnum = 0;
> +	*completed = 0;
> +}
>  static inline void rcutorture_record_test_transition(void)
>  {
>  }
> diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
> index bd30bc61bc05..1110db210318 100644
> --- a/kernel/rcu/rcutorture.c
> +++ b/kernel/rcu/rcutorture.c
> @@ -138,6 +138,15 @@ static long n_barrier_attempts;
>  static long n_barrier_successes;
>  static struct list_head rcu_torture_removed;
>  
> +static int rcu_torture_writer_state;
> +#define RTWS_FIXED_DELAY	0
> +#define RTWS_DELAY		1
> +#define RTWS_REPLACE		2
> +#define RTWS_DEF_FREE		3
> +#define RTWS_EXP_SYNC		4
> +#define RTWS_STUTTER		5
> +#define RTWS_STOPPING		6
> +
>  #if defined(MODULE) || defined(CONFIG_RCU_TORTURE_TEST_RUNNABLE)
>  #define RCUTORTURE_RUNNABLE_INIT 1
>  #else
> @@ -214,6 +223,7 @@ rcu_torture_free(struct rcu_torture *p)
>   */
>  
>  struct rcu_torture_ops {
> +	int ttype;
>  	void (*init)(void);
>  	int (*readlock)(void);
>  	void (*read_delay)(struct torture_random_state *rrsp);
> @@ -312,6 +322,7 @@ static void rcu_sync_torture_init(void)
>  }
>  
>  static struct rcu_torture_ops rcu_ops = {
> +	.ttype		= RTORT_RCU,
>  	.init		= rcu_sync_torture_init,
>  	.readlock	= rcu_torture_read_lock,
>  	.read_delay	= rcu_read_delay,
> @@ -355,6 +366,7 @@ static void rcu_bh_torture_deferred_free(struct rcu_torture *p)
>  }
>  
>  static struct rcu_torture_ops rcu_bh_ops = {
> +	.ttype		= RTORT_RCU_BH,
>  	.init		= rcu_sync_torture_init,
>  	.readlock	= rcu_bh_torture_read_lock,
>  	.read_delay	= rcu_read_delay,  /* just reuse rcu's version. */
> @@ -397,6 +409,7 @@ call_rcu_busted(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
>  }
>  
>  static struct rcu_torture_ops rcu_busted_ops = {
> +	.ttype		= RTORT_BUSTED,
>  	.init		= rcu_sync_torture_init,
>  	.readlock	= rcu_torture_read_lock,
>  	.read_delay	= rcu_read_delay,  /* just reuse rcu's version. */
> @@ -492,6 +505,7 @@ static void srcu_torture_synchronize_expedited(void)
>  }
>  
>  static struct rcu_torture_ops srcu_ops = {
> +	.ttype		= RTORT_SRCU,
>  	.init		= rcu_sync_torture_init,
>  	.readlock	= srcu_torture_read_lock,
>  	.read_delay	= srcu_read_delay,
> @@ -527,6 +541,7 @@ static void rcu_sched_torture_deferred_free(struct rcu_torture *p)
>  }
>  
>  static struct rcu_torture_ops sched_ops = {
> +	.ttype		= RTORT_RCU_SCHED,
>  	.init		= rcu_sync_torture_init,
>  	.readlock	= sched_torture_read_lock,
>  	.read_delay	= rcu_read_delay,  /* just reuse rcu's version. */
> @@ -699,12 +714,15 @@ rcu_torture_writer(void *arg)
>  	set_user_nice(current, MAX_NICE);
>  
>  	do {
> +		rcu_torture_writer_state = RTWS_FIXED_DELAY;
>  		schedule_timeout_uninterruptible(1);
>  		rp = rcu_torture_alloc();
>  		if (rp == NULL)
>  			continue;
>  		rp->rtort_pipe_count = 0;
> +		rcu_torture_writer_state = RTWS_DELAY;
>  		udelay(torture_random(&rand) & 0x3ff);
> +		rcu_torture_writer_state = RTWS_REPLACE;
>  		old_rp = rcu_dereference_check(rcu_torture_current,
>  					       current == writer_task);
>  		rp->rtort_mbtest = 1;
> @@ -721,8 +739,10 @@ rcu_torture_writer(void *arg)
>  			else
>  				exp = gp_exp;
>  			if (!exp) {
> +				rcu_torture_writer_state = RTWS_DEF_FREE;
>  				cur_ops->deferred_free(old_rp);
>  			} else {
> +				rcu_torture_writer_state = RTWS_EXP_SYNC;
>  				cur_ops->exp_sync();
>  				list_add(&old_rp->rtort_free,
>  					 &rcu_torture_removed);
> @@ -743,8 +763,10 @@ rcu_torture_writer(void *arg)
>  			}
>  		}
>  		rcutorture_record_progress(++rcu_torture_current_version);
> +		rcu_torture_writer_state = RTWS_STUTTER;
>  		stutter_wait("rcu_torture_writer");
>  	} while (!torture_must_stop());
> +	rcu_torture_writer_state = RTWS_STOPPING;
>  	torture_kthread_stopping("rcu_torture_writer");
>  	return 0;
>  }
> @@ -937,6 +959,7 @@ rcu_torture_printk(char *page)
>  	int i;
>  	long pipesummary[RCU_TORTURE_PIPE_LEN + 1] = { 0 };
>  	long batchsummary[RCU_TORTURE_PIPE_LEN + 1] = { 0 };
> +	static unsigned long rtcv_snap = ULONG_MAX;
>  
>  	for_each_possible_cpu(cpu) {
>  		for (i = 0; i < RCU_TORTURE_PIPE_LEN + 1; i++) {
> @@ -997,6 +1020,20 @@ rcu_torture_printk(char *page)
>  	page += sprintf(page, "\n");
>  	if (cur_ops->stats)
>  		cur_ops->stats(page);
> +	if (rtcv_snap == rcu_torture_current_version &&
> +	    rcu_torture_current != NULL) {
> +		int __maybe_unused flags;
> +		unsigned long __maybe_unused gpnum;
> +		unsigned long __maybe_unused completed;
> +
> +		rcutorture_get_gp_data(cur_ops->ttype,
> +				       &flags, &gpnum, &completed);
> +		page += sprintf(page,
> +				"??? Writer stall state %d g%lu c%lu f%#x\n",
> +				rcu_torture_writer_state,
> +				gpnum, completed, flags);
> +	}
> +	rtcv_snap = rcu_torture_current_version;
>  }
>  
>  /*
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index 0c47e300210a..032106df7391 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -294,6 +294,24 @@ void rcutorture_record_test_transition(void)
>  EXPORT_SYMBOL_GPL(rcutorture_record_test_transition);
>  
>  /*
> + * Send along grace-period-related data for rcutorture diagnostics.
> + */
> +void rcutorture_get_gp_data(enum rcutorture_type test_type, int *flags,
> +			    unsigned long *gpnum, unsigned long *completed)
> +{
> +	if (test_type == RTORT_SRCU || test_type == RTORT_BUSTED) {
> +		*flags = 0;
> +		*gpnum = 0;
> +		*completed = 0;
> +		return;
> +	}
> +	*flags = ACCESS_ONCE(rcu_state->gp_flags);
> +	*gpnum = ACCESS_ONCE(rcu_state->gpnum);
> +	*completed = ACCESS_ONCE(rcu_state->completed);
> +}
> +EXPORT_SYMBOL_GPL(rcutorture_get_gp_data);
> +
> +/*
>   * Record the number of writer passes through the current rcutorture test.
>   * This is also used to correlate debugfs tracing stats with the rcutorture
>   * messages.
> -- 
> 1.8.1.5
>