From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 112794 invoked by alias); 17 Dec 2019 21:48:04 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Received: (qmail 112106 invoked by uid 89); 17 Dec 2019 21:48:00 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-21.9 required=5.0 tests=AWL,BAYES_00,GIT_PATCH_0,GIT_PATCH_1,GIT_PATCH_2,GIT_PATCH_3,RCVD_IN_DNSWL_NONE,SPF_PASS autolearn=ham version=3.3.1 spammy=HX-Languages-Length:2092 X-HELO: mail-pj1-f65.google.com DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=from:to:subject:date:message-id:in-reply-to:references; bh=EbVlLte6cM3D+rEIFJmkDeBnjV3lHOp6fJKNaMCCY8s=; b=Fwkmz7HYFhKXJpiNY5gDRdsMRBfSO6Vxg33ZvHf8KH9JMqBL3t/RgbMc8FL176NbCg MKSDluHrcdJ1u18KPFaz8mQzTit0AbWo/4hQwz1yu2mO9+81gFIaMyF5t178sEymojWo 2ud2BJNZ/ibu+w3WRfdE0PQiMIbFv16NPGNuhEUYFLyOPgVdgIny3lyPM5VlTlKINPsG UZHC7gipe3XOU3AXnamIZE4qAl6WTLtoAKTpGWPQ0Ua5wn4P3fLMdkT9YHy+DfQSj5ub xnU/kuKrpGgCFIfPekWP66WgMfAN5Vsr33KKigZUnf38lrhUtEG1+9rFPphgXJwMhmA7 w3FQ== Return-Path: From: Adhemerval Zanella To: libc-alpha@sourceware.org Subject: [PATCH 16/16] linux: Optimize fallback 32-bit clock_getres Date: Tue, 17 Dec 2019 21:48:00 -0000 Message-Id: <20191217214728.2886-16-adhemerval.zanella@linaro.org> In-Reply-To: <20191217214728.2886-1-adhemerval.zanella@linaro.org> References: <20191217214728.2886-1-adhemerval.zanella@linaro.org> X-SW-Source: 2019-12/txt/msg00602.txt.bz2 This patch avoid probing the __NR_clock_getttime64 syscall each time __clock_gettime64 is issued on a kernel without 64 bit time support. Once ENOSYS is obtained, only 32-bit clock_gettime are used. The following snippet: clock_getres (CLOCK_REALTIME, &(struct timespec) { 0 }); clock_getres (CLOCK_MONOTONIC, &(struct timespec) { 0 }); clock_getres (CLOCK_BOOTTIME, &(struct timespec) { 0 }); clock_getres (20, &(struct timespec) { 0 }); On a kernel without 64 bit time support issues the syscalls: syscall_0x196(0, 0xffb83330, [...]) = -1 ENOSYS (Function not implemented) clock_getres(CLOCK_REALTIME, {tv_sec=0, tv_nsec=1}) = 0 clock_getres(CLOCK_MONOTONIC, {tv_sec=0, tv_nsec=1}) = 0 clock_getres(CLOCK_BOOTTIME, {tv_sec=0, tv_nsec=1}) = 0 Checked on i686-linux-gnu on 4.15 kernel. --- sysdeps/unix/sysv/linux/clock_getres.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) diff --git a/sysdeps/unix/sysv/linux/clock_getres.c b/sysdeps/unix/sysv/linux/clock_getres.c index 3d59f11d71..c1e1cf60c6 100644 --- a/sysdeps/unix/sysv/linux/clock_getres.c +++ b/sysdeps/unix/sysv/linux/clock_getres.c @@ -42,13 +42,21 @@ __clock_getres64 (clockid_t clock_id, struct __timespec64 *res) int r; /* Old 32-bit ABI with possible 64-bit time_t support. */ # ifdef __NR_clock_getres_time64 + /* Avoid issue a __NR_clock_getres_time64 syscall on kernels that do not + support 64-bit time_t. */ + static int time64_support = 1; + if (atomic_load_relaxed (&time64_support) == 1) + { # ifdef HAVE_CLOCK_GETRES64_VSYSCALL - r = INLINE_VSYSCALL (clock_getres_time64, 2, clock_id, res); + r = INLINE_VSYSCALL (clock_getres_time64, 2, clock_id, res); # else - r = INLINE_SYSCALL_CALL (clock_getres_time64, clock_id, res); + r = INLINE_SYSCALL_CALL (clock_getres_time64, clock_id, res); # endif - if (r == 0 || errno != ENOSYS) - return r; + if (r == 0 || errno != ENOSYS) + return r; + + atomic_store_relaxed (&time64_support, 0); + } # endif /* Fallback code that uses 32-bit support. */ struct timespec ts32; -- 2.17.1