Re: [PATCH V2 5/6] x86/intel_rdt: Use perf infrastructure for measurements

[Peter Zijlstra]

On Thu, Aug 16, 2018 at 01:16:08PM -0700, Reinette Chatre wrote:
> +	l2_miss_event = perf_event_create_kernel_counter(&perf_miss_attr,
> +							 plr->cpu,
> +							 NULL, NULL, NULL);
> +	if (IS_ERR(l2_miss_event))
> +		goto out;
> +
> +	l2_hit_event = perf_event_create_kernel_counter(&perf_hit_attr,
> +							plr->cpu,
> +							NULL, NULL, NULL);
> +	if (IS_ERR(l2_hit_event))
> +		goto out_l2_miss;
> +
> +	local_irq_disable();
> +	/*
> +	 * Check any possible error state of events used by performing
> +	 * one local read.
> +	 */
> +	if (perf_event_read_local(l2_miss_event, &tmp, NULL, NULL)) {
> +		local_irq_enable();
> +		goto out_l2_hit;
> +	}
> +	if (perf_event_read_local(l2_hit_event, &tmp, NULL, NULL)) {
> +		local_irq_enable();
> +		goto out_l2_hit;
> +	}
> +
> +	/*
> +	 * Disable hardware prefetchers.
>  	 *
> +	 * Call wrmsr direcly to avoid the local register variables from
> +	 * being overwritten due to reordering of their assignment with
> +	 * the wrmsr calls.
> +	 */
> +	__wrmsr(MSR_MISC_FEATURE_CONTROL, prefetch_disable_bits, 0x0);
> +
> +	/* Initialize rest of local variables */
> +	/*
> +	 * Performance event has been validated right before this with
> +	 * interrupts disabled - it is thus safe to read the counter index.
> +	 */
> +	l2_miss_pmcnum = x86_perf_rdpmc_ctr_get(l2_miss_event);
> +	l2_hit_pmcnum = x86_perf_rdpmc_ctr_get(l2_hit_event);
> +	line_size = plr->line_size;
> +	mem_r = plr->kmem;
> +	size = plr->size;
> +
> +	/*
> +	 * Read counter variables twice - first to load the instructions
> +	 * used in L1 cache, second to capture accurate value that does not
> +	 * include cache misses incurred because of instruction loads.
> +	 */
> +	rdpmcl(l2_hit_pmcnum, l2_hits_before);
> +	rdpmcl(l2_miss_pmcnum, l2_miss_before);
> +	/*
> +	 * From SDM: Performing back-to-back fast reads are not guaranteed
> +	 * to be monotonic. To guarantee monotonicity on back-toback reads,
> +	 * a serializing instruction must be placed between the two
> +	 * RDPMC instructions
> +	 */
> +	rmb();
> +	rdpmcl(l2_hit_pmcnum, l2_hits_before);
> +	rdpmcl(l2_miss_pmcnum, l2_miss_before);
> +	/*
> +	 * rdpmc is not a serializing instruction. Add barrier to prevent
> +	 * instructions that follow to begin executing before reading the
> +	 * counter value.
> +	 */
> +	rmb();
> +	for (i = 0; i < size; i += line_size) {
> +		/*
> +		 * Add a barrier to prevent speculative execution of this
> +		 * loop reading beyond the end of the buffer.
> +		 */
> +		rmb();
> +		asm volatile("mov (%0,%1,1), %%eax\n\t"
> +			     :
> +			     : "r" (mem_r), "r" (i)
> +			     : "%eax", "memory");
> +	}
> +	rdpmcl(l2_hit_pmcnum, l2_hits_after);
> +	rdpmcl(l2_miss_pmcnum, l2_miss_after);
> +	/*
> +	 * rdpmc is not a serializing instruction. Add barrier to ensure
> +	 * events measured have completed and prevent instructions that
> +	 * follow to begin executing before reading the counter value.
> +	 */
> +	rmb();
> +	/* Re-enable hardware prefetchers */
> +	wrmsr(MSR_MISC_FEATURE_CONTROL, 0x0, 0x0);
> +	local_irq_enable();
> +	trace_pseudo_lock_l2(l2_hits_after - l2_hits_before,
> +			     l2_miss_after - l2_miss_before);
> +
> +out_l2_hit:
> +	perf_event_release_kernel(l2_hit_event);
> +out_l2_miss:
> +	perf_event_release_kernel(l2_miss_event);
> +out:
> +	plr->thread_done = 1;
> +	wake_up_interruptible(&plr->lock_thread_wq);
> +	return 0;
> +}
> +

The above, looks a _LOT_ like the below. And while C does suck a little,
I'm sure there's something we can do about this.

> +	l3_miss_event = perf_event_create_kernel_counter(&perf_miss_attr,
> +							 plr->cpu,
> +							 NULL, NULL,
> +							 NULL);
> +	if (IS_ERR(l3_miss_event))
> +		goto out;
> +
> +	l3_hit_event = perf_event_create_kernel_counter(&perf_hit_attr,
> +							plr->cpu,
> +							NULL, NULL,
> +							NULL);
> +	if (IS_ERR(l3_hit_event))
> +		goto out_l3_miss;
> +
>  	local_irq_disable();
>  	/*
> +	 * Check any possible error state of events used by performing
> +	 * one local read.
> +	 */
> +	if (perf_event_read_local(l3_miss_event, &tmp, NULL, NULL)) {
> +		local_irq_enable();
> +		goto out_l3_hit;
> +	}
> +	if (perf_event_read_local(l3_hit_event, &tmp, NULL, NULL)) {
> +		local_irq_enable();
> +		goto out_l3_hit;
> +	}
> +
> +	/*
> +	 * Disable hardware prefetchers.
> +	 *
>  	 * Call wrmsr direcly to avoid the local register variables from
>  	 * being overwritten due to reordering of their assignment with
>  	 * the wrmsr calls.
>  	 */
>  	__wrmsr(MSR_MISC_FEATURE_CONTROL, prefetch_disable_bits, 0x0);
> +
> +	/* Initialize rest of local variables */
> +	/*
> +	 * Performance event has been validated right before this with
> +	 * interrupts disabled - it is thus safe to read the counter index.
> +	 */
> +	l3_hit_pmcnum = x86_perf_rdpmc_ctr_get(l3_hit_event);
> +	l3_miss_pmcnum = x86_perf_rdpmc_ctr_get(l3_miss_event);
> +	line_size = plr->line_size;
>  	mem_r = plr->kmem;
>  	size = plr->size;
> +
> +	/*
> +	 * Read counter variables twice - first to load the instructions
> +	 * used in L1 cache, second to capture accurate value that does not
> +	 * include cache misses incurred because of instruction loads.
> +	 */
> +	rdpmcl(l3_hit_pmcnum, l3_hits_before);
> +	rdpmcl(l3_miss_pmcnum, l3_miss_before);
> +	/*
> +	 * From SDM: Performing back-to-back fast reads are not guaranteed
> +	 * to be monotonic. To guarantee monotonicity on back-toback reads,
> +	 * a serializing instruction must be placed between the two
> +	 * RDPMC instructions
> +	 */
> +	rmb();
> +	rdpmcl(l3_hit_pmcnum, l3_hits_before);
> +	rdpmcl(l3_miss_pmcnum, l3_miss_before);
> +	/*
> +	 * rdpmc is not a serializing instruction. Add barrier to prevent
> +	 * instructions that follow to begin executing before reading the
> +	 * counter value.
> +	 */
> +	rmb();
>  	for (i = 0; i < size; i += line_size) {
> +		/*
> +		 * Add a barrier to prevent speculative execution of this
> +		 * loop reading beyond the end of the buffer.
> +		 */
> +		rmb();
>  		asm volatile("mov (%0,%1,1), %%eax\n\t"
>  			     :
>  			     : "r" (mem_r), "r" (i)
>  			     : "%eax", "memory");
>  	}
> +	rdpmcl(l3_hit_pmcnum, l3_hits_after);
> +	rdpmcl(l3_miss_pmcnum, l3_miss_after);
>  	/*
> +	 * rdpmc is not a serializing instruction. Add barrier to ensure
> +	 * events measured have completed and prevent instructions that
> +	 * follow to begin executing before reading the counter value.
>  	 */
> +	rmb();
> +	/* Re-enable hardware prefetchers */
>  	wrmsr(MSR_MISC_FEATURE_CONTROL, 0x0, 0x0);
>  	local_irq_enable();
> +	l3_miss_after -= l3_miss_before;
> +	if (boot_cpu_data.x86_model == INTEL_FAM6_BROADWELL_X) {
> +		/*
> +		 * On BDW references and misses are counted, need to adjust.
> +		 * Sometimes the "hits" counter is a bit more than the
> +		 * references, for example, x references but x + 1 hits.
> +		 * To not report invalid hit values in this case we treat
> +		 * that as misses equal to references.
> +		 */
> +		/* First compute the number of cache references measured */
> +		l3_hits_after -= l3_hits_before;
> +		/* Next convert references to cache hits */
> +		l3_hits_after -= l3_miss_after > l3_hits_after ?
> +					l3_hits_after : l3_miss_after;
> +	} else {
> +		l3_hits_after -= l3_hits_before;
>  	}
> +	trace_pseudo_lock_l3(l3_hits_after, l3_miss_after);
>  
> +out_l3_hit:
> +	perf_event_release_kernel(l3_hit_event);
> +out_l3_miss:
> +	perf_event_release_kernel(l3_miss_event);