From patchwork Mon Apr 25 13:01:50 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Adhemerval Zanella Netto X-Patchwork-Id: 53177 Return-Path: X-Original-To: patchwork@sourceware.org Delivered-To: patchwork@sourceware.org Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id 41887385781D for ; Mon, 25 Apr 2022 13:04:00 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 41887385781D DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sourceware.org; s=default; t=1650891840; bh=X0G9gQOl9UXbLfKyK+ehCnNF2N7pSc7zExFSl7FdpR8=; h=To:Subject:Date:In-Reply-To:References:List-Id:List-Unsubscribe: List-Archive:List-Post:List-Help:List-Subscribe:From:Reply-To: From; b=RqGWJwUEDcv6T0KNZXclU5+lZGTIY3ZNDTFEI7IzTaZqUMwmRWZEptKc9lXVK+GsP lo45wrF3b0KtpKQCHoOdw9MStHCupl1jyccFZrZ/3mOM4RwCA69j0HMf4iu9tOidVj tw5rG/APVC/Xu2znHVa6avlv+/l0zXyUgQdXNQBE= X-Original-To: libc-alpha@sourceware.org Delivered-To: libc-alpha@sourceware.org Received: from mail-oi1-x233.google.com (mail-oi1-x233.google.com [IPv6:2607:f8b0:4864:20::233]) by sourceware.org (Postfix) with ESMTPS id 566383858C52 for ; Mon, 25 Apr 2022 13:02:13 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 566383858C52 Received: by mail-oi1-x233.google.com with SMTP id t15so17004706oie.1 for ; Mon, 25 Apr 2022 06:02:13 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=X0G9gQOl9UXbLfKyK+ehCnNF2N7pSc7zExFSl7FdpR8=; b=Pt9EuBmv+w50MTmFdPx6pHQGBeIBSVJH2CNfouv8qy28LTT8Qvmd2GUcv+4LFfYEkb TQdo1wiyrxWtfW71q7IJ/yZ14C+gfsVpLCx7/2eXCqjQ/yIym3ClMTeukGMcKYPwlqYB vtTLr0bDPDrqM68Ykymc4LfcQRbpvEBMRVS+jTygUG1XEON5BFX96d9VRHcMDoltsFpG JJfKrd7TuKbVvrEDpBoUtq8Mm/1BNfpjG3awm/fnBK7YlETC5KLrLGzry9ZH0K272Lak +KRs4dDrhPF4ddyD4LbZsYGpSxwK5q5uKVaNUNEG6XIL1HQCQtFQif2lqiQadNOXKMto i+7g== X-Gm-Message-State: AOAM532uoZMHJOQmCPY70vegm8pkgSVKHrk4WTkoZUl2j7jD2Z4y2tci Iscqj3OgUIvroIHZfRx4Wi+GOW9SyASURQ== X-Google-Smtp-Source: ABdhPJxxsMH85j1YojMOzEXOG9OeqjzYBIDuW9AW5vL2hT/6DVWTTVVhQ/j69LhBggLLjbStX74hMA== X-Received: by 2002:a05:6808:2199:b0:322:b431:e240 with SMTP id be25-20020a056808219900b00322b431e240mr12757863oib.118.1650891731834; Mon, 25 Apr 2022 06:02:11 -0700 (PDT) Received: from birita.. ([2804:431:c7ca:4214:b4dd:3339:98d6:1ec0]) by smtp.gmail.com with ESMTPSA id e4-20020a0568301e4400b006054dca5442sm3804278otj.28.2022.04.25.06.02.09 for (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 25 Apr 2022 06:02:10 -0700 (PDT) To: libc-alpha@sourceware.org Subject: [PATCH v4 3/9] benchtests: Add arc4random benchtest Date: Mon, 25 Apr 2022 10:01:50 -0300 Message-Id: <20220425130156.1062525-4-adhemerval.zanella@linaro.org> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20220425130156.1062525-1-adhemerval.zanella@linaro.org> References: <20220425130156.1062525-1-adhemerval.zanella@linaro.org> MIME-Version: 1.0 X-Spam-Status: No, score=-12.2 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, GIT_PATCH_0, KAM_SHORT, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-Patchwork-Original-From: Adhemerval Zanella via Libc-alpha From: Adhemerval Zanella Netto Reply-To: Adhemerval Zanella Errors-To: libc-alpha-bounces+patchwork=sourceware.org@sourceware.org Sender: "Libc-alpha" It shows both throughput (total bytes obtained in the test duration) and latecy for both arc4random and arc4random_buf with different sizes. Checked on x86_64-linux-gnu, aarch64-linux, and powerpc64le-linux-gnu. --- NEWS | 6 +- benchtests/Makefile | 6 +- benchtests/bench-arc4random.c | 224 ++++++++++++++++++++++++++++++++++ stdlib/arc4random.c | 23 ++-- 4 files changed, 243 insertions(+), 16 deletions(-) create mode 100644 benchtests/bench-arc4random.c diff --git a/NEWS b/NEWS index 4d9d95b35b..57e2fb2c3b 100644 --- a/NEWS +++ b/NEWS @@ -9,9 +9,9 @@ Version 2.36 Major new features: -* The functions arc4random, arc4random_buf, arc4random_uniform have been - added. The functions use a cryptographic pseudo-random number generator - based on ChaCha20 initilized with entropy from kernel. +* The functions arc4random, arc4random_buf, and arc4random_uniform have been + added. These functions use a cryptographic pseudo-random number generator + based on ChaCha20 initialized with entropy from the kernel. Deprecated and removed features, and other changes affecting compatibility: diff --git a/benchtests/Makefile b/benchtests/Makefile index 8dfca592fd..50b96dd71f 100644 --- a/benchtests/Makefile +++ b/benchtests/Makefile @@ -111,8 +111,12 @@ bench-string := \ ffsll \ # bench-string +bench-stdlib := \ + arc4random \ +# bench-stdlib + ifeq (${BENCHSET},) -bench := $(bench-math) $(bench-pthread) $(bench-string) +bench := $(bench-math) $(bench-pthread) $(bench-string) $(bench-stdlib) else bench := $(foreach B,$(filter bench-%,${BENCHSET}), ${${B}}) endif diff --git a/benchtests/bench-arc4random.c b/benchtests/bench-arc4random.c new file mode 100644 index 0000000000..626f2ba48c --- /dev/null +++ b/benchtests/bench-arc4random.c @@ -0,0 +1,224 @@ +/* arc4random benchmarks. + Copyright (C) 2022 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + . */ + +#include "bench-timing.h" +#include "json-lib.h" +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Prevent compiler to optimize away call. */ +#define DO_NOT_OPTIMIZE(__value) \ + ({ \ + __typeof (__value) __v = (__value); \ + asm volatile("" : : "r,m" (__v) : "memory");\ + }) + +static volatile sig_atomic_t timer_finished; + +static void timer_callback (int unused) +{ + timer_finished = 1; +} + +static timer_t timer; + +/* Run for approximately DURATION seconds, and it does not matter who + receive the signal (so not need to mask it on main thread). */ +static void +timer_start (void) +{ + timer_finished = 0; + timer = support_create_timer (DURATION, 0, false, timer_callback); +} +static void +timer_stop (void) +{ + support_delete_timer (timer); +} + +static const uint32_t sizes[] = { 0, 16, 32, 48, 64, 80, 96, 112, 128 }; + +static double +bench_throughput (void) +{ + uint64_t n = 0; + + struct timespec start, end; + clock_gettime (CLOCK_MONOTONIC, &start); + while (1) + { + DO_NOT_OPTIMIZE (arc4random ()); + n++; + + if (timer_finished == 1) + break; + } + clock_gettime (CLOCK_MONOTONIC, &end); + struct timespec diff = timespec_sub (end, start); + + double total = (double) n * sizeof (uint32_t); + double duration = (double) diff.tv_sec + + (double) diff.tv_nsec / TIMESPEC_HZ; + + return total / duration; +} + +static double +bench_latency (void) +{ + timing_t start, stop, cur; + const size_t iters = 1024; + + TIMING_NOW (start); + for (size_t i = 0; i < iters; i++) + DO_NOT_OPTIMIZE (arc4random ()); + TIMING_NOW (stop); + + TIMING_DIFF (cur, start, stop); + + return (double) (cur) / (double) iters; +} + +static double +bench_buf_throughput (size_t len) +{ + uint8_t buf[len]; + uint64_t n = 0; + + struct timespec start, end; + clock_gettime (CLOCK_MONOTONIC, &start); + while (1) + { + arc4random_buf (buf, len); + n++; + + if (timer_finished == 1) + break; + } + clock_gettime (CLOCK_MONOTONIC, &end); + struct timespec diff = timespec_sub (end, start); + + double total = (double) n * len; + double duration = (double) diff.tv_sec + + (double) diff.tv_nsec / TIMESPEC_HZ; + + return total / duration; +} + +static double +bench_buf_latency (size_t len) +{ + timing_t start, stop, cur; + const size_t iters = 1024; + + uint8_t buf[len]; + + TIMING_NOW (start); + for (size_t i = 0; i < iters; i++) + arc4random_buf (buf, len); + TIMING_NOW (stop); + + TIMING_DIFF (cur, start, stop); + + return (double) (cur) / (double) iters; +} + +static void +bench_singlethread (json_ctx_t *json_ctx) +{ + json_element_object_begin (json_ctx); + + json_array_begin (json_ctx, "throughput"); + for (int i = 0; i < array_length (sizes); i++) + { + timer_start (); + double r = sizes[i] == 0 + ? bench_throughput () : bench_buf_throughput (sizes[i]); + timer_stop (); + + json_element_double (json_ctx, r); + } + json_array_end (json_ctx); + + json_array_begin (json_ctx, "latency"); + for (int i = 0; i < array_length (sizes); i++) + { + timer_start (); + double r = sizes[i] == 0 + ? bench_latency () : bench_buf_latency (sizes[i]); + timer_stop (); + + json_element_double (json_ctx, r); + } + json_array_end (json_ctx); + + json_element_object_end (json_ctx); +} + +static void +run_bench (json_ctx_t *json_ctx, const char *name, + char *const*fnames, size_t fnameslen, + void (*bench)(json_ctx_t *ctx)) +{ + json_attr_object_begin (json_ctx, name); + json_array_begin (json_ctx, "functions"); + for (int i = 0; i < fnameslen; i++) + json_element_string (json_ctx, fnames[i]); + json_array_end (json_ctx); + + json_array_begin (json_ctx, "results"); + bench (json_ctx); + json_array_end (json_ctx); + json_attr_object_end (json_ctx); +} + +static int +do_test (void) +{ + char *fnames[array_length (sizes)]; + for (int i = 0; i < array_length (sizes); i++) + if (sizes[i] == 0) + fnames[i] = xasprintf ("arc4random"); + else + fnames[i] = xasprintf ("arc4random_buf(%u)", sizes[i]); + + json_ctx_t json_ctx; + json_init (&json_ctx, 0, stdout); + + json_document_begin (&json_ctx); + json_attr_string (&json_ctx, "timing_type", TIMING_TYPE); + + run_bench (&json_ctx, "single-thread", fnames, array_length (fnames), + bench_singlethread); + + json_document_end (&json_ctx); + + for (int i = 0; i < array_length (sizes); i++) + free (fnames[i]); + + return 0; +} + +#include diff --git a/stdlib/arc4random.c b/stdlib/arc4random.c index c42dd0b1e1..19a486405e 100644 --- a/stdlib/arc4random.c +++ b/stdlib/arc4random.c @@ -25,17 +25,16 @@ #include #include -/* Besides the cipher state 'ctx', it keeps two counters: 'have' is the - current valid bytes not yet consumed in 'buf', while 'count' is the maximum - number of bytes until a reseed. +/* arc4random keeps two counters: 'have' is the current valid bytes not yet + consumed in 'buf' while 'count' is the maximum number of bytes until a + reseed. Both the initial seed and reseed tries to obtain entropy from the kernel and abort the process if none could be obtained. - The state 'buf' improves the usage of the cipher call, allowing to call - optimized implementations (if the archictecture provides it) and optimize - arc4random calls (since only multiple call it will encrypt the next block). - */ + A large 'buf' also improves the cipher throughput and allows to call of + optimized implementations (where it usually operates on multiple blocks). +*/ /* Maximum number bytes until reseed (16 MB). */ #define CHACHA20_RESEED_SIZE (16 * 1024 * 1024) @@ -63,7 +62,7 @@ __libc_lock_define_initialized (, __arc4random_lock); void __arc4random_fork_subprocess (void) { - if (__arc4random_wipeonfork && state != NULL) + if (!__arc4random_wipeonfork && state != NULL) memset (state, 0, sizeof (struct arc4random_state)); __libc_lock_init (__arc4random_lock); @@ -81,10 +80,10 @@ arc4random_getrandom_failure (void) __libc_fatal ("Fatal glibc error: Cannot get entropy for arc4random\n"); } -/* Fork detection is done by checking if MADV_WIPEONFORK supported. If not - the fork callback will reset the state on the fork call. It does not - handle direct clone calls, nor vfork or _Fork (arc4random is not - async-signal-safe due the internal lock usage). */ +/* If the kernel supports MADV_WIPEONFORK it is used to provide fork + detection. Otherwise, the state is resetied with an atfork handler. The + latter does not handle direct clone calls, nor vfork or _Fork (arc4random + is not async-signal-safe due to the internal lock usage). */ static void arc4random_init (uint8_t *buf) {