This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 00/12] Y2038 preparation: use clock_[gs]ettime to implement the other time-getting and -setting functions
- From: Alistair Francis <alistair23 at gmail dot com>
- To: Zack Weinberg <zackw at panix dot com>
- Cc: GNU C Library <libc-alpha at sourceware dot org>, Joseph Myers <joseph at codesourcery dot com>, Florian Weimer <fweimer at redhat dot com>, Lukasz Majewski <lukma at denx dot de>, Stepan Golosunov <stepan at golosunov dot pp dot ru>, Arnd Bergmann <arnd at arndb dot de>
- Date: Tue, 20 Aug 2019 11:43:12 -0700
- Subject: Re: [PATCH 00/12] Y2038 preparation: use clock_[gs]ettime to implement the other time-getting and -setting functions
- References: <20190820132152.24100-1-zackw@panix.com>
On Tue, Aug 20, 2019 at 6:21 AM Zack Weinberg <zackw@panix.com> wrote:
>
> This patchset aims to make the Y2038 project a little easier by
> implementing the other time-getting and time-setting functions
> (time, ftime, stime, gettimeofday, and settimeofday) in terms of
> clock_gettime and clock_settime. Internal uses of (__)gettimeofday
> are also all changed to __clock_gettime. Internal uses of time() are
> mostly left alone, since time() is easier and clearer when you don’t
> care about sub-second resolution and don’t need a struct timespec for
> some other reason.
Thanks for the series, this should help. I'll rebase my work ontop of
this once it's merged into master.
>
> As a consequence, gettimeofday will no longer report crude timezone
> information under any circumstances. If its ‘tzp’ argument is not
> NULL, it will write zeroes to both fields of the ‘struct timezone’,
> even if the gettimeofday system call exists and would have produced
> different output. Similarly, ftime will always write zeroes to the
> timezone and dstflag fields of ‘struct timeb’. Joseph Myers raised an
> objection to this change; I rebutted the objection, and he did not
> reply; I do not think either of us will persuade the other, so I’m
> formally asking for a third maintainer to make a decision.
I like this and think it makes a lot of sense.
Alistair
>
> settimeofday _will_ still call the settimeofday system call and pass
> down a non-null ‘tzp’ argument if it receives one, but, if _both_ of
> its arguments are non-null, it will fail and set errno to EINVAL.
> Also, its invocation of a settimeofday system call goes through an
> internal sysdep function called __settimezone. On the Hurd this is an
> unconditional ENOSYS stub, and on Linux it’s an ENOSYS stub when
> __NR_settimeofday is not defined.
>
> Relatedly, there is a macro-based hack to issue warnings when
> gettimeofday is called with a ‘tzp’ argument that is not a
> compile-time null pointer. I think this is probably a better option
> at this stage than various suggested alternatives (e.g. renaming the
> fields of ‘struct timezone’, or removing it from sys/time.h
> altogether) because the remaining valid uses of settimeofday with
> non-null ‘tzp’ should still be able to use struct timezone. The hack
> is in a separate patch because, well, it’s a hack and I’m hoping
> someone has a better implementation idea. (I tried an inline function
> instead of a macro, but it threw false positives all over the place on
> ‘gettimeofday (&tv, 0)’...)
>
> This patchset also includes an overhaul of Linux/Alpha’s GLIBC_2.0
> compatibility symbols for several system calls that consume or produce
> struct timeval. It appears that early releases of glibc and/or Linux
> for Alpha attempted to achieve some degree of binary compatibility
> with OSF/1, but this came at the expense of using 32-bit time_t inside
> struct timeval. Up till now, we were invoking compatibility system
> calls from the compatibility symbols. This patch changes them to wrap
> the 64-bit implementations and convert between 32-bit and 64-bit
> struct timeval in user space. If a 64-bit tv_sec value returned by
> the kernel is found to be greater than INT32_MAX, the caller will
> receive a saturated 32-bit struct timeval, { INT32_MAX, 0 }, and the
> function will return -1 with errno set to EOVERFLOW. Because many of
> these calls are assumed unable to fail, we make sure to write all the
> fields of any output data structure even when this happens. I did
> this partially because it simplified conversion of gettimeofday to
> call clock_gettime, and partially as a worked example of the semantics
> that I think are probably best for the 32-bit time_t compatibility
> symbols we’re going to have to add for many other architectures.
> Comments on these semantics are requested.
>
> This patchset also includes a partial revision of manual/time.texi.
> clock_gettime and clock_settime are now documented, and the
> obsolescent status of stime, settimeofday, and gettimeofday is
> clearer. I only documented CLOCK_REALTIME and CLOCK_MONOTONIC,
> because most of the other clock constants are either extremely
> Linux-specific or they have to do with measuring CPU time, and I
> didn’t touch the measuring-CPU-time part of time.texi. (That part
> also deserves a rewrite, but I don’t know enough about the topic and
> this patchset is already long enough.)
>
> Finally, there is a case for demoting ftime and stime to compatibility
> symbols and removing their public prototypes (for ftime, we would stop
> installing <sys/timeb.h> altogether). I tried to investigate how much
> this might break, using codesearch.debian.net, but the identifiers
> ‘ftime’ and ‘stime’ turn out to be heavily used for unrelated
> purposes, so I didn’t get anywhere.
>
> Zack Weinberg (12):
> Remove implementations of clock_[gs]ettime using [gs]ettimeofday.
> Change most internal uses of __gettimeofday to __clock_gettime.
> Don’t use the argument to time.
> Use clock_settime to implement stime.
> Use clock_settime to implement settimeofday.
> Use clock_gettime to implement time.
> Use clock_gettime to implement ftime.
> Use clock_gettime to implement gettimeofday.
> Finish move of clock_* functions to libc.
> Warn when gettimeofday is called with non-null tzp argument.
> Linux/Alpha: don’t use timeval32 system calls.
> Revise the documentation of ‘simple calendar time’.
>
> NEWS | 53 +-
> include/sys/time.h | 7 +-
> include/time.h | 4 +-
> inet/deadline.c | 9 +-
> login/logout.c | 9 +-
> login/logwtmp.c | 7 +-
> manual/filesys.texi | 2 +-
> manual/llio.texi | 10 +-
> manual/threads.texi | 2 +-
> manual/time.texi | 1062 +++++++++--------
> misc/syslog.c | 2 +-
> nis/nis_call.c | 4 +-
> nptl/pthread_join_common.c | 7 +-
> nptl/pthread_mutex_timedlock.c | 7 +-
> nscd/nscd_helper.c | 24 +-
> resolv/gai_misc.c | 6 +-
> resolv/gai_suspend.c | 6 +-
> resolv/res_send.c | 6 +-
> rt/Makefile | 15 +-
> rt/Versions | 7 +-
> rt/clock-compat.c | 63 -
> string/strfry.c | 2 +-
> sunrpc/auth_des.c | 19 +-
> sunrpc/auth_unix.c | 9 +-
> sunrpc/create_xid.c | 6 +-
> sunrpc/svcauth_des.c | 7 +-
> support/support_test_main.c | 14 +-
> sysdeps/generic/memusage.h | 16 +-
> .../mach/{gettimeofday.c => clock_gettime.c} | 25 +-
> .../hurd/{settimeofday.c => clock_settime.c} | 27 +-
> sysdeps/mach/hurd/getitimer.c | 3 +-
> sysdeps/mach/hurd/i386/libc.abilist | 1 -
> sysdeps/mach/hurd/i386/librt.abilist | 5 -
> sysdeps/mach/hurd/setitimer.c | 3 +-
> sysdeps/mach/hurd/times.c | 6 +-
> sysdeps/mach/nanosleep.c | 33 +-
> sysdeps/mach/sleep.c | 4 +-
> sysdeps/mach/usleep.c | 5 -
> sysdeps/posix/gettimeofday.c | 67 --
> sysdeps/posix/tempname.c | 9 +-
> sysdeps/posix/time.c | 40 -
> sysdeps/pthread/aio_misc.c | 6 +-
> sysdeps/pthread/aio_suspend.c | 6 +-
> sysdeps/unix/clock_gettime.c | 64 -
> sysdeps/unix/make-syscalls.sh | 2 +-
> sysdeps/unix/syscalls.list | 2 -
> .../unix/sysv/linux/aarch64/gettimeofday.c | 71 --
> sysdeps/unix/sysv/linux/aarch64/init-first.c | 7 +-
> sysdeps/unix/sysv/linux/aarch64/libc-vdso.h | 2 -
> sysdeps/unix/sysv/linux/aarch64/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/alpha/Makefile | 2 +-
> sysdeps/unix/sysv/linux/alpha/Versions | 1 -
> sysdeps/unix/sysv/linux/alpha/adjtime.c | 70 +-
> .../linux/{time.c => alpha/gettimeofday.c} | 31 +-
> sysdeps/unix/sysv/linux/alpha/libc.abilist | 2 -
> sysdeps/unix/sysv/linux/alpha/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/alpha/osf_adjtime.c | 135 +++
> sysdeps/unix/sysv/linux/alpha/osf_getitimer.c | 48 +
> sysdeps/unix/sysv/linux/alpha/osf_getrusage.c | 44 +
> .../unix/sysv/linux/alpha/osf_gettimeofday.c | 53 +
> sysdeps/unix/sysv/linux/alpha/osf_setitimer.c | 55 +
> .../unix/sysv/linux/alpha/osf_settimeofday.c | 51 +
> .../ftime.c => sysv/linux/alpha/osf_utimes.c} | 34 +-
> .../linux/alpha/osf_wait4.c} | 51 +-
> sysdeps/unix/sysv/linux/alpha/settimeofday.c | 22 +
> sysdeps/unix/sysv/linux/alpha/syscalls.list | 13 +-
> sysdeps/unix/sysv/linux/alpha/tv32-compat.h | 151 +++
> sysdeps/unix/sysv/linux/arm/init-first.c | 7 +-
> sysdeps/unix/sysv/linux/arm/libc-vdso.h | 2 -
> sysdeps/unix/sysv/linux/arm/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/arm/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/clock_settime.c | 1 +
> sysdeps/unix/sysv/linux/csky/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/ftime.c | 3 -
> sysdeps/unix/sysv/linux/gettimeofday.c | 39 -
> sysdeps/unix/sysv/linux/hppa/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/hppa/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/i386/gettimeofday.c | 35 -
> sysdeps/unix/sysv/linux/i386/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/i386/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/i386/time.c | 34 -
> sysdeps/unix/sysv/linux/ia64/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/ia64/librt.abilist | 5 -
> .../sysv/linux/m68k/coldfire/libc.abilist | 1 -
> .../sysv/linux/m68k/coldfire/librt.abilist | 5 -
> .../unix/sysv/linux/m68k/m680x0/libc.abilist | 1 -
> .../unix/sysv/linux/m68k/m680x0/librt.abilist | 5 -
> .../unix/sysv/linux/microblaze/libc.abilist | 1 -
> .../unix/sysv/linux/microblaze/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/mips/init-first.c | 7 +-
> sysdeps/unix/sysv/linux/mips/libc-vdso.h | 2 -
> .../sysv/linux/mips/mips32/fpu/libc.abilist | 1 -
> .../unix/sysv/linux/mips/mips32/librt.abilist | 5 -
> .../sysv/linux/mips/mips32/nofpu/libc.abilist | 1 -
> .../unix/sysv/linux/mips/mips64/librt.abilist | 5 -
> .../sysv/linux/mips/mips64/n32/libc.abilist | 1 -
> .../sysv/linux/mips/mips64/n64/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/nios2/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/powerpc/Versions | 1 -
> .../unix/sysv/linux/powerpc/gettimeofday.c | 85 --
> sysdeps/unix/sysv/linux/powerpc/init-first.c | 13 +-
> sysdeps/unix/sysv/linux/powerpc/libc-vdso.h | 2 -
> .../linux/powerpc/powerpc32/fpu/libc.abilist | 1 -
> .../linux/powerpc/powerpc32/librt.abilist | 5 -
> .../powerpc/powerpc32/nofpu/libc.abilist | 1 -
> .../linux/powerpc/powerpc64/be/libc.abilist | 1 -
> .../linux/powerpc/powerpc64/be/librt.abilist | 5 -
> .../linux/powerpc/powerpc64/le/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/powerpc/time.c | 84 --
> sysdeps/unix/sysv/linux/riscv/init-first.c | 10 +-
> sysdeps/unix/sysv/linux/riscv/libc-vdso.h | 2 -
> .../unix/sysv/linux/riscv/rv64/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/s390/init-first.c | 9 +-
> sysdeps/unix/sysv/linux/s390/libc-vdso.h | 3 -
> .../unix/sysv/linux/s390/s390-32/libc.abilist | 1 -
> .../sysv/linux/s390/s390-32/librt.abilist | 5 -
> .../unix/sysv/linux/s390/s390-64/libc.abilist | 1 -
> .../sysv/linux/s390/s390-64/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/settimezone.c | 39 +
> sysdeps/unix/sysv/linux/sh/libc.abilist | 1 -
> sysdeps/unix/sysv/linux/sh/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/sparc/init-first.c | 8 +-
> sysdeps/unix/sysv/linux/sparc/libc-vdso.h | 2 -
> .../sysv/linux/sparc/sparc32/libc.abilist | 1 -
> .../sysv/linux/sparc/sparc32/librt.abilist | 5 -
> .../sysv/linux/sparc/sparc64/libc.abilist | 1 -
> .../sysv/linux/sparc/sparc64/librt.abilist | 5 -
> sysdeps/unix/sysv/linux/sparc/sparc64/time.c | 1 -
> sysdeps/unix/sysv/linux/syscalls.list | 1 -
> sysdeps/unix/sysv/linux/x86/gettimeofday.c | 61 -
> sysdeps/unix/sysv/linux/x86/time.c | 59 -
> .../unix/sysv/linux/x86_64/64/libc.abilist | 1 -
> .../unix/sysv/linux/x86_64/64/librt.abilist | 5 -
> .../unix/sysv/linux/x86_64/x32/libc.abilist | 1 -
> .../unix/sysv/linux/x86_64/x32/librt.abilist | 5 -
> .../unix/sysv/linux/x86_64/x32/syscalls.list | 2 -
> time/Makefile | 8 +-
> time/Versions | 2 +-
> {rt => time}/clock_getcpuclockid.c | 0
> {rt => time}/clock_getres.c | 0
> {rt => time}/clock_gettime.c | 0
> {rt => time}/clock_nanosleep.c | 0
> {rt => time}/clock_settime.c | 1 +
> time/ftime.c | 26 +-
> time/getdate.c | 2 +-
> time/gettimeofday.c | 30 +-
> time/settimeofday.c | 24 +-
> sysdeps/unix/stime.c => time/settimezone.c | 23 +-
> time/stime.c | 15 +-
> time/sys/time.h | 36 +-
> time/time.c | 17 +-
> {rt => time}/tst-clock.c | 0
> {rt => time}/tst-clock2.c | 0
> {rt => time}/tst-clock_nanosleep.c | 0
> {rt => time}/tst-cpuclock1.c | 0
> time/tst_wcsftime.c | 2 +-
> 156 files changed, 1580 insertions(+), 1775 deletions(-)
> delete mode 100644 rt/clock-compat.c
> rename sysdeps/mach/{gettimeofday.c => clock_gettime.c} (67%)
> rename sysdeps/mach/hurd/{settimeofday.c => clock_settime.c} (71%)
> delete mode 100644 sysdeps/posix/gettimeofday.c
> delete mode 100644 sysdeps/posix/time.c
> delete mode 100644 sysdeps/unix/clock_gettime.c
> delete mode 100644 sysdeps/unix/sysv/linux/aarch64/gettimeofday.c
> rename sysdeps/unix/sysv/linux/{time.c => alpha/gettimeofday.c} (64%)
> create mode 100644 sysdeps/unix/sysv/linux/alpha/osf_adjtime.c
> create mode 100644 sysdeps/unix/sysv/linux/alpha/osf_getitimer.c
> create mode 100644 sysdeps/unix/sysv/linux/alpha/osf_getrusage.c
> create mode 100644 sysdeps/unix/sysv/linux/alpha/osf_gettimeofday.c
> create mode 100644 sysdeps/unix/sysv/linux/alpha/osf_setitimer.c
> create mode 100644 sysdeps/unix/sysv/linux/alpha/osf_settimeofday.c
> rename sysdeps/unix/{bsd/ftime.c => sysv/linux/alpha/osf_utimes.c} (60%)
> rename sysdeps/unix/{clock_settime.c => sysv/linux/alpha/osf_wait4.c} (55%)
> create mode 100644 sysdeps/unix/sysv/linux/alpha/settimeofday.c
> create mode 100644 sysdeps/unix/sysv/linux/alpha/tv32-compat.h
> delete mode 100644 sysdeps/unix/sysv/linux/ftime.c
> delete mode 100644 sysdeps/unix/sysv/linux/gettimeofday.c
> delete mode 100644 sysdeps/unix/sysv/linux/i386/gettimeofday.c
> delete mode 100644 sysdeps/unix/sysv/linux/i386/time.c
> delete mode 100644 sysdeps/unix/sysv/linux/powerpc/gettimeofday.c
> delete mode 100644 sysdeps/unix/sysv/linux/powerpc/time.c
> create mode 100644 sysdeps/unix/sysv/linux/settimezone.c
> delete mode 100644 sysdeps/unix/sysv/linux/sparc/sparc64/time.c
> delete mode 100644 sysdeps/unix/sysv/linux/x86/gettimeofday.c
> delete mode 100644 sysdeps/unix/sysv/linux/x86/time.c
> rename {rt => time}/clock_getcpuclockid.c (100%)
> rename {rt => time}/clock_getres.c (100%)
> rename {rt => time}/clock_gettime.c (100%)
> rename {rt => time}/clock_nanosleep.c (100%)
> rename {rt => time}/clock_settime.c (96%)
> rename sysdeps/unix/stime.c => time/settimezone.c (67%)
> rename {rt => time}/tst-clock.c (100%)
> rename {rt => time}/tst-clock2.c (100%)
> rename {rt => time}/tst-clock_nanosleep.c (100%)
> rename {rt => time}/tst-cpuclock1.c (100%)
>
> --
> 2.23.0.rc1
>