linux-kernel - [RFC 5/6] perf/core: rotation no longer necessary. Behavior has changed. Beware

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]

Date:   Tue, 10 Jan 2017 02:25:01 -0800
From:   David Carrillo-Cisneros <davidcc@...gle.com>
To:     linux-kernel@...r.kernel.org
Cc:     "x86@...nel.org" <x86@...nel.org>, Ingo Molnar <mingo@...hat.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Andi Kleen <ak@...ux.intel.com>,
        Kan Liang <kan.liang@...el.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Borislav Petkov <bp@...e.de>,
        Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        Vikas Shivappa <vikas.shivappa@...ux.intel.com>,
        Mark Rutland <mark.rutland@....com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Vince Weaver <vince@...ter.net>, Paul Turner <pjt@...gle.com>,
        Stephane Eranian <eranian@...gle.com>,
        David Carrillo-Cisneros <davidcc@...gle.com>
Subject: [RFC 5/6] perf/core: rotation no longer necessary. Behavior has changed. Beware

The sched in/out process updates timestamps and "rotates"
ctx->inactive_groups.

This changes the speed at which rotation happens. Before events will
rotate one event per interruption, now they will rotate q events each
timer interruption. Where q is the number of events added to the pmu per
sched in.

Signed-off-by: David Carrillo-Cisneros <davidcc@...gle.com>
---
 kernel/events/core.c | 22 +++++-----------------
 1 file changed, 5 insertions(+), 17 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index c7715b2627a9..f5d9c13b485f 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -3642,19 +3642,6 @@ static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx,
 	raw_spin_unlock(&ctx->lock);
 }
 
-/*
- * Round-robin a context's events:
- */
-static void rotate_ctx(struct perf_event_context *ctx)
-{
-	/*
-	 * Rotate the first entry last of non-pinned groups. Rotation might be
-	 * disabled by the inheritance code.
-	 */
-	if (!ctx->rotate_disable)
-		list_rotate_left(&ctx->flexible_groups);
-}
-
 static int perf_rotate_context(struct perf_cpu_context *cpuctx)
 {
 	struct perf_event_context *ctx = NULL;
@@ -3681,10 +3668,11 @@ static int perf_rotate_context(struct perf_cpu_context *cpuctx)
 	if (ctx)
 		ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE);
 
-	rotate_ctx(&cpuctx->ctx);
-	if (ctx)
-		rotate_ctx(ctx);
-
+	/*
+	 * A sched out will insert event groups at end of inactive_groups,
+	 * a sched in will schedule events at the beginning of inactive_groups.
+	 * This causes a rotation.
+	 */
 	perf_event_sched_in(cpuctx, ctx, current);
 
 	perf_pmu_enable(cpuctx->ctx.pmu);
-- 
2.11.0.390.gc69c2f50cf-goog