linux-kernel - [PATCH v6 4/6] perf: Add per perf_cpu_context min

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20200214075133.181299-5-irogers@google.com>
Date:   Thu, 13 Feb 2020 23:51:31 -0800
From:   Ian Rogers <irogers@...gle.com>
To:     Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Jiri Olsa <jolsa@...hat.com>,
        Namhyung Kim <namhyung@...nel.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Randy Dunlap <rdunlap@...radead.org>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>,
        Shuah Khan <skhan@...uxfoundation.org>,
        Krzysztof Kozlowski <krzk@...nel.org>,
        Kees Cook <keescook@...omium.org>,
        "Paul E. McKenney" <paulmck@...nel.org>,
        Masami Hiramatsu <mhiramat@...nel.org>,
        Marco Elver <elver@...gle.com>,
        Kent Overstreet <kent.overstreet@...il.com>,
        Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
        Ard Biesheuvel <ardb@...nel.org>,
        Gary Hook <Gary.Hook@....com>,
        Kan Liang <kan.liang@...ux.intel.com>,
        linux-kernel@...r.kernel.org
Cc:     Stephane Eranian <eranian@...gle.com>,
        Andi Kleen <ak@...ux.intel.com>,
        Ian Rogers <irogers@...gle.com>
Subject: [PATCH v6 4/6] perf: Add per perf_cpu_context min_heap storage

The storage required for visit_groups_merge's min heap needs to vary in
order to support more iterators, such as when multiple nested cgroups'
events are being visited. This change allows for 2 iterators and doesn't
support growth.
Based-on-work-by: Peter Zijlstra (Intel) <peterz@...radead.org>

Signed-off-by: Ian Rogers <irogers@...gle.com>
---
 include/linux/perf_event.h |  7 +++++
 kernel/events/core.c       | 53 ++++++++++++++++++++++++++------------
 2 files changed, 44 insertions(+), 16 deletions(-)

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 68e21e828893..5060e31b32cc 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -862,6 +862,13 @@ struct perf_cpu_context {
 	int				sched_cb_usage;
 
 	int				online;
+	/*
+	 * Per-CPU storage for iterators used in visit_groups_merge. The default
+	 * storage is of size 2 to hold the CPU and any CPU event iterators.
+	 */
+	int				itr_storage_cap;
+	struct perf_event		**itr_storage;
+	struct perf_event		*itr_default[2];
 };
 
 struct perf_output_handle {
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 832e2a56a663..18e4bb871d85 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -3419,36 +3419,48 @@ static void __heap_add(struct min_heap *heap, struct perf_event *event)
 	}
 }
 
-static noinline int visit_groups_merge(struct perf_event_groups *groups,
-				int cpu,
+static noinline int visit_groups_merge(struct perf_cpu_context *cpuctx,
+				struct perf_event_groups *groups, int cpu,
 				int (*func)(struct perf_event *, void *),
 				void *data)
 {
 	/* Space for per CPU and/or any CPU event iterators. */
 	struct perf_event *itrs[2];
-	struct min_heap event_heap = {
-		.data = itrs,
-		.size = 0,
-		.cap = ARRAY_SIZE(itrs),
-	};
+	struct min_heap event_heap;
+	struct perf_event **evt;
 	struct perf_event *next;
 	int ret;
 
-	__heap_add(&event_heap, perf_event_groups_first(groups, -1));
+	if (cpuctx) {
+		event_heap = (struct min_heap){
+			.data = cpuctx->itr_storage,
+			.size = 0,
+			.cap = cpuctx->itr_storage_cap,
+		};
+	} else {
+		event_heap = (struct min_heap){
+			.data = itrs,
+			.size = 0,
+			.cap = ARRAY_SIZE(itrs),
+		};
+		/* Events not within a CPU context may be on any CPU. */
+		__heap_add(&event_heap, perf_event_groups_first(groups, -1));
+	}
+	evt = event_heap.data;
+
 	__heap_add(&event_heap, perf_event_groups_first(groups, cpu));
 
 	min_heapify_all(&event_heap, &perf_min_heap);
 
 	while (event_heap.size) {
-		ret = func(itrs[0], data);
+		ret = func(*evt, data);
 		if (ret)
 			return ret;
 
-		next = perf_event_groups_next(itrs[0]);
-		if (next) {
-			min_heap_pop_push(&event_heap, &next,
-					&perf_min_heap);
-		} else
+		next = perf_event_groups_next(*evt);
+		if (next)
+			min_heap_pop_push(&event_heap, &next, &perf_min_heap);
+		else
 			min_heap_pop(&event_heap, &perf_min_heap);
 	}
 
@@ -3519,7 +3531,10 @@ ctx_pinned_sched_in(struct perf_event_context *ctx,
 		.can_add_hw = 1,
 	};
 
-	visit_groups_merge(&ctx->pinned_groups,
+	if (ctx != &cpuctx->ctx)
+		cpuctx = NULL;
+
+	visit_groups_merge(cpuctx, &ctx->pinned_groups,
 			   smp_processor_id(),
 			   pinned_sched_in, &sid);
 }
@@ -3534,7 +3549,10 @@ ctx_flexible_sched_in(struct perf_event_context *ctx,
 		.can_add_hw = 1,
 	};
 
-	visit_groups_merge(&ctx->flexible_groups,
+	if (ctx != &cpuctx->ctx)
+		cpuctx = NULL;
+
+	visit_groups_merge(cpuctx, &ctx->flexible_groups,
 			   smp_processor_id(),
 			   flexible_sched_in, &sid);
 }
@@ -10395,6 +10413,9 @@ int perf_pmu_register(struct pmu *pmu, const char *name, int type)
 		cpuctx->online = cpumask_test_cpu(cpu, perf_online_mask);
 
 		__perf_mux_hrtimer_init(cpuctx, cpu);
+
+		cpuctx->itr_storage_cap = ARRAY_SIZE(cpuctx->itr_default);
+		cpuctx->itr_storage = cpuctx->itr_default;
 	}
 
 got_cpu_context:
-- 
2.25.0.265.gbab2e86ba0-goog