Re: [RFC PATCH 1/1] smp: Add tracepoints for functions called with smp_call_function*()

[Leonardo Brás]

On Thu, 2023-05-04 at 12:59 +0100, Valentin Schneider wrote:
> +Daniel
> 
> On 03/05/23 16:59, Peter Zijlstra wrote:
> > On Wed, Apr 19, 2023 at 12:45:08AM -0300, Leonardo Brás wrote:
> > > 
> > > IIUC the last commits add tracepoints that are collected in the
> > > requesting CPU, at the moment of scheduling the IPI, which are also useful in
> > > some scenarios.
> > > 
> > > On my scenario, it could help a little, since it makes possible to filter what
> > > all other cpus are scheduling on the requested cpu. OTOH it could be also be
> > > misleading, as the requested cpu could be running something that was scheduled
> > > way before.
> > > 
> > > The change I propose does exactly what my scenario need: track exactly which
> > > function was running at given time in the requested CPU. With this info, we can
> > > check which (if any) remotely requested function was running on given time
> > > window.
> > 
> > I was thinking you could simply (graph)-trace
> > __flush_smp_call_function_queue() with a max_graph_depth or so (Steve
> > says that ought to work).
> > 
> > But even that might be too specific, your use case sounds more like what
> > we have the irq-off latency tracer for, and that thing will immediately
> > tell you what functions were being ran.
> > 
> > > (An unrelated thing I just thought: We could even use the commits you pointed
> > > together with my proposed change in order to measure how long does it take for a
> > > requested function to run / complete in the requested cpu)
> > 
> > I don't think you could actually do that; the send tracepoints Valentin
> > added don't log the csd address, this means you cannot distinguish
> > two CSDs with the same function send from different CPUs.
> > 
> > To do this you'd need to add the csd address to the the ipi_send
> > tracepoints and your own (possibly replacing info -- because I don't
> > think that's too useful).
> > 
> > Valentin -- is any of this something you'd also find useful?
> 
> Conceptually yeah, however:
> 
> With isolcpus + NOHZ_FULL, *any* IPI sent to an isolated CPU is game over,
> you interrupt the userspace task and you eat the full meal that is
> NOHZ_FULL kernel entry. Pretty much any such IPI will show up when
> evaluating your setup with rteval/rlta/whatever, so having an event at
> IPI reception is redundant.
> 
> IIUC Leonardo's looking at a setup with isolcpus but not necessarily
> NOHZ_FULL, so he wants to look at *which* IPIs (if any) are pushing the
> isolated task over its deadline/period. I would argue any IPI received
> there is at risk of doing that, so it's the same fight as with NOHZ_FULL.
> 
> With that said, I suppose this could still be helpful for e.g. osnoise to
> hook into and point the finger at which CPU/context sent the problematic
> IPI. Or more generally, as Leonardo suggested, to measure CSD IPI delivery
> times.
> 
> One thing though is that trace_ipi_send_cpu*() is not used solely for
> CSD's, cf. irq_work_raise() or smp_send_reschedule(). We might want to
> split that into e.g. trace_csd_queue_cpu*() + trace_ipi_send*().
> 
> Something like so...
> 
> ---
> diff --git a/include/trace/events/smp.h b/include/trace/events/smp.h
> new file mode 100644
> index 0000000000000..8fc725a2b45b9
> --- /dev/null
> +++ b/include/trace/events/smp.h
> @@ -0,0 +1,69 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +#undef TRACE_SYSTEM
> +#define TRACE_SYSTEM smp
> +
> +#if !defined(_TRACE_SMP_H) || defined(TRACE_HEADER_MULTI_READ)
> +#define _TRACE_SMP_H
> +
> +#include <linux/tracepoint.h>
> +
> +TRACE_EVENT(csd_queue_cpu,
> +
> +	TP_PROTO(const unsigned int cpu,
> +		 unsigned long callsite,
> +		 smp_call_func_t func,
> +		 call_single_data_t *csd),
> +
> +	TP_ARGS(cpu, callsite, func, csd),
> +
> +	TP_STRUCT__entry(
> +		__field(unsigned int, cpu)
> +		__field(void *, callsite)
> +		__field(void *, func)
> +		__field(void *, csd)
> +	),
> +
> +	TP_fast_assign(
> +		__entry->cpu = cpu;
> +		__entry->callsite = (void *)callsite;
> +		__entry->func = func;
> +		__entry->csd  = csd;
> +	),
> +
> +	TP_printk("cpu=%u callsite=%pS func=%pS csd=%p",
> +		  __entry->cpu, __entry->callsite, __entry->func, __entry->csd)
> +);

This is for the caller side, right?

> +
> +DECLARE_EVENT_CLASS(csd_function,
> +
> +	TP_PROTO(smp_call_func_t func, call_single_data_t *csd),
> +
> +	TP_ARGS(func, csd),
> +
> +	TP_STRUCT__entry(
> +		__field(void *,	func)
> +		__field(void *,	csd)
> +	),
> +
> +	TP_fast_assign(
> +		__entry->func	= func;
> +		__entry->csd	= csd;
> +	),
> +
> +	TP_printk("func=%pS csd=%p", __entry->func, __entry->csd)
> +);
> +
> +DEFINE_EVENT(csd_function, csd_function_entry,
> +	TP_PROTO(smp_call_func_t func, call_single_data_t *csd),
> +	TP_ARGS(func, csd)
> +);
> +
> +DEFINE_EVENT(csd_function, csd_function_exit,
> +	TP_PROTO(smp_call_func_t func, call_single_data_t *csd),
> +	TP_ARGS(func, csd)
> +);

Oh, this is what event_class is for. Thanks for the example :)

> +
> +#endif /* _TRACE_SMP_H */
> +
> +/* This part must be outside protection */
> +#include <trace/define_trace.h>
> diff --git a/kernel/smp.c b/kernel/smp.c
> index ab3e5dad6cfe9..7d28db303e9bc 100644
> --- a/kernel/smp.c
> +++ b/kernel/smp.c
> @@ -27,6 +27,9 @@
>  #include <linux/jump_label.h>
>  
>  #include <trace/events/ipi.h>
> +#define CREATE_TRACE_POINTS
> +#include <trace/events/smp.h>
> +#undef CREATE_TRACE_POINTS
>  
>  #include "smpboot.h"
>  #include "sched/smp.h"
> @@ -121,6 +124,14 @@ send_call_function_ipi_mask(struct cpumask *mask)
>  	arch_send_call_function_ipi_mask(mask);
>  }
>  
> +static __always_inline void
> +csd_do_func(smp_call_func_t func, void *info, call_single_data_t *csd)
> +{
> +	trace_csd_function_entry(func, csd);
> +	func(info);
> +	trace_csd_function_exit(func, csd);
> +}
> +

Good one, a helper to avoid calling those traces everywhere. 

Honest question: 
Since info == csd->info and func == csd->func, we could just pass csd, right?
I suppose the suggestion on the 3-argument version is to use the values already
fetched from memory instead of fetching them again. Is that correct?

>  #ifdef CONFIG_CSD_LOCK_WAIT_DEBUG
>  
>  static DEFINE_STATIC_KEY_MAYBE(CONFIG_CSD_LOCK_WAIT_DEBUG_DEFAULT, csdlock_debug_enabled);
> @@ -329,7 +340,7 @@ void __smp_call_single_queue(int cpu, struct llist_node *node)
>  	 * even if we haven't sent the smp_call IPI yet (e.g. the stopper
>  	 * executes migration_cpu_stop() on the remote CPU).
>  	 */
> -	if (trace_ipi_send_cpu_enabled()) {
> +	if (trace_csd_queue_cpu_enabled()) {
>  		call_single_data_t *csd;
>  		smp_call_func_t func;
>  
> @@ -337,7 +348,7 @@ void __smp_call_single_queue(int cpu, struct llist_node *node)
>  		func = CSD_TYPE(csd) == CSD_TYPE_TTWU ?
>  			sched_ttwu_pending : csd->func;
>  
> -		trace_ipi_send_cpu(cpu, _RET_IP_, func);
> +		trace_csd_queue_cpu(cpu, _RET_IP_, func, csd);
>  	}
>  
>  	/*
> @@ -375,7 +386,7 @@ static int generic_exec_single(int cpu, struct __call_single_data *csd)
>  		csd_lock_record(csd);
>  		csd_unlock(csd);
>  		local_irq_save(flags);
> -		func(info);
> +		csd_do_func(func, info, NULL);
>  		csd_lock_record(NULL);
>  		local_irq_restore(flags);
>  		return 0;
> @@ -477,7 +488,7 @@ static void __flush_smp_call_function_queue(bool warn_cpu_offline)
>  			}
>  
>  			csd_lock_record(csd);
> -			func(info);
> +			csd_do_func(func, info, csd);
>  			csd_unlock(csd);
>  			csd_lock_record(NULL);
>  		} else {
> @@ -508,7 +519,7 @@ static void __flush_smp_call_function_queue(bool warn_cpu_offline)
>  
>  				csd_lock_record(csd);
>  				csd_unlock(csd);
> -				func(info);
> +				csd_do_func(func, info, csd);
>  				csd_lock_record(NULL);
>  			} else if (type == CSD_TYPE_IRQ_WORK) {
>  				irq_work_single(csd);
> @@ -522,8 +533,10 @@ static void __flush_smp_call_function_queue(bool warn_cpu_offline)
>  	/*
>  	 * Third; only CSD_TYPE_TTWU is left, issue those.
>  	 */
> -	if (entry)
> -		sched_ttwu_pending(entry);
> +	if (entry) {
> +		csd = llist_entry(entry, typeof(*csd), node.llist);
> +		csd_do_func(sched_ttwu_pending, entry, csd);
> +	}
>  }
>  
>  
> @@ -624,7 +637,7 @@ EXPORT_SYMBOL(smp_call_function_single);
>  
>  /**
>   * smp_call_function_single_async() - Run an asynchronous function on a
> - * 			         specific CPU.
> + *				 specific CPU.
>   * @cpu: The CPU to run on.
>   * @csd: Pre-allocated and setup data structure
>   *
> @@ -728,7 +741,7 @@ static void smp_call_function_many_cond(const struct cpumask *mask,
>  	int cpu, last_cpu, this_cpu = smp_processor_id();
>  	struct call_function_data *cfd;
>  	bool wait = scf_flags & SCF_WAIT;
> -	int nr_cpus = 0, nr_queued = 0;
> +	int nr_cpus = 0;
>  	bool run_remote = false;
>  	bool run_local = false;
>  
> @@ -786,21 +799,16 @@ static void smp_call_function_many_cond(const struct cpumask *mask,
>  			csd->node.src = smp_processor_id();
>  			csd->node.dst = cpu;
>  #endif
> +
> +			trace_csd_queue_cpu(cpu, _RET_IP_, func, csd);
> +
>  			if (llist_add(&csd->node.llist, &per_cpu(call_single_queue, cpu))) {
>  				__cpumask_set_cpu(cpu, cfd->cpumask_ipi);
>  				nr_cpus++;
>  				last_cpu = cpu;
>  			}
> -			nr_queued++;
>  		}
>  
> -		/*
> -		 * Trace each smp_function_call_*() as an IPI, actual IPIs
> -		 * will be traced with func==generic_smp_call_function_single_ipi().
> -		 */
> -		if (nr_queued)
> -			trace_ipi_send_cpumask(cfd->cpumask, _RET_IP_, func);
> -
>  		/*
>  		 * Choose the most efficient way to send an IPI. Note that the
>  		 * number of CPUs might be zero due to concurrent changes to the
> @@ -816,7 +824,7 @@ static void smp_call_function_many_cond(const struct cpumask *mask,
>  		unsigned long flags;
>  
>  		local_irq_save(flags);
> -		func(info);
> +		csd_do_func(func, info, NULL);
>  		local_irq_restore(flags);
>  	}
>  
> 

I will rebase my patch on top of tip/smp/core and apply the suggested changes.
A v2 will be send shortly.

Thank you all for reviewing!
Leo