From patchwork Wed Aug 28 18:31:22 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Zack Weinberg X-Patchwork-Id: 34314 Received: (qmail 115793 invoked by alias); 28 Aug 2019 18:31:38 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 115781 invoked by uid 89); 28 Aug 2019 18:31:38 -0000 Authentication-Results: sourceware.org; auth=none X-Spam-SWARE-Status: No, score=-14.6 required=5.0 tests=AWL, BAYES_00, GIT_PATCH_0, GIT_PATCH_1, GIT_PATCH_2, GIT_PATCH_3, SPF_PASS autolearn=ham version=3.3.1 spammy=HX-Received:ea16 X-HELO: mailbackend.panix.com MIME-Version: 1.0 References: <20190827173015.24370-1-alistair.francis@wdc.com> In-Reply-To: From: Zack Weinberg Date: Wed, 28 Aug 2019 14:31:22 -0400 Message-ID: Subject: Re: [PATCH v9] y2038: Introduce the __ASSUME_TIME64_SYSCALLS define To: Joseph Myers Cc: Alistair Francis , Alistair Francis , GNU C Library , Lukasz Majewski OK, how does this look? From dc644211be404aea73b6bd218777c6f06472d9f1 Mon Sep 17 00:00:00 2001 From: Lukasz Majewski Date: Wed, 28 Aug 2019 14:27:03 -0400 Subject: [PATCH] y2038: Introduce the __ASSUME_TIME64_SYSCALLS define Add a macro to linux/kernel-features.h, __ASSUME_TIME64_SYSCALLS, to indicate whether the kernel can be assumed to provide a set of system calls that process 64-bit time_t. __ASSUME_TIME64_SYSCALLS does not indicate whether time_t is actually 64 bits (that's __TIMEBITS) and also does not indicate whether the 64-bit time_t system calls have "time64" suffixes on their names. Code that uses __ASSUME_TIME64_SYSCALLS will be added in subsequent patches. * sysdeps/unix/sysv/linux/kernel-features.h (__ASSUME_TIME64_SYSCALLS): New macro. --- sysdeps/unix/sysv/linux/kernel-features.h | 59 +++++++++++++++++++++++ 1 file changed, 59 insertions(+) diff --git a/sysdeps/unix/sysv/linux/kernel-features.h b/sysdeps/unix/sysv/linux/kernel-features.h index 1518bb5228..179df79433 100644 --- a/sysdeps/unix/sysv/linux/kernel-features.h +++ b/sysdeps/unix/sysv/linux/kernel-features.h @@ -20,6 +20,11 @@ /* This file must not contain any C code. At least it must be protected to allow using the file also in assembler files. */ +#ifndef _LINUX_KERNEL_FEATURES_H +#define _LINUX_KERNEL_FEATURES_H 1 + +#include + #ifndef __LINUX_KERNEL_VERSION /* We assume the worst; all kernels should be supported. */ # define __LINUX_KERNEL_VERSION 0 @@ -139,3 +144,57 @@ */ #define __ASSUME_CLONE_DEFAULT 1 + +/* Support for 64-bit time_t in the system call interface. When this + flag is set, the kernel provides a version of each of these system + calls that accepts 64-bit time_t: + + clock_adjtime(64) + clock_gettime(64) + clock_settime(64) + clock_getres(_time64) + clock_nanosleep(_time64) + futex(_time64) + mq_timedreceive(_time64) + mq_timedsend(_time64) + ppoll(_time64) + pselect6(_time64) + rt_sigtimedwait(_time64) + sched_rr_get_interval(_time64) + timer_gettime(64) + timer_settime(64) + timerfd_gettime(64) + timerfd_settime(64) + utimensat(_time64) + + On architectures where time_t has historically been 64 bits, + only the 64-bit version of each system call exists, and there + are no suffixes on the __NR_ constants. + + On architectures where time_t has historically been 32 bits, + both 32-bit and 64-bit versions of each system call may exist, + depending on the kernel version. When the 64-bit version exists, + there is a '64' or '_time64' suffix on the name of its __NR_ + constant, as shown above. + + This flag is always set for Linux 5.1 and later. Prior to that + version, it is set only for some CPU architectures and ABIs: + + - __WORDSIZE == 64 - all supported architectures where pointers + are 64 bits also have always had 64-bit time_t. + + - __WORDSIZE == 32 && __SYSCALL_WORDSIZE == 64 - this describes + only one supported configuration, x86's 'x32' subarchitecture, + where pointers are 32 bits but time_t has always been 64 bits. + + __ASSUME_TIME64_SYSCALLS being set does not mean __TIMESIZE is 64, + and __TIMESIZE equal to 64 does not mean __ASSUME_TIME64_SYSCALLS + is set. All four cases are possible. */ + +#if __LINUX_KERNEL_VERSION >= 0x050100 \ + || __WORDSIZE == 64 \ + || (defined __SYSCALL_WORDSIZE && __SYSCALL_WORDSIZE == 64) +# define __ASSUME_TIME64_SYSCALLS 1 +#endif + +#endif /* kernel-features.h */