time-util: deal with systems where userspace has 64bit time_t but kernel does not

Fixes: #14362
This commit is contained in:
Lennart Poettering 2019-12-17 11:22:17 +01:00 committed by Yu Watanabe
parent 13b6c4c8de
commit 601f91bec5

View file

@ -1500,8 +1500,29 @@ int time_change_fd(void) {
if (fd < 0)
return -errno;
if (timerfd_settime(fd, TFD_TIMER_ABSTIME|TFD_TIMER_CANCEL_ON_SET, &its, NULL) < 0)
return -errno;
if (timerfd_settime(fd, TFD_TIMER_ABSTIME|TFD_TIMER_CANCEL_ON_SET, &its, NULL) >= 0)
return TAKE_FD(fd);
return TAKE_FD(fd);
/* So apparently there are systems where time_t is 64bit, but the kernel actually doesn't support
* 64bit time_t. In that case configuring a timer to TIME_T_MAX will fail with EOPNOTSUPP or a
* similar error. If that's the case let's try with INT32_MAX instead, maybe that works. It's a bit
* of a black magic thing though, but what can we do?
*
* We don't want this code on x86-64, hence let's conditionalize this for systems with 64bit time_t
* but where "long" is shorter than 64bit, i.e. 32bit archs.
*
* See: https://github.com/systemd/systemd/issues/14362 */
#if SIZEOF_TIME_T == 8 && ULONG_MAX < UINT64_MAX
if (ERRNO_IS_NOT_SUPPORTED(errno)) {
static const struct itimerspec its32 = {
.it_value.tv_sec = INT32_MAX,
};
if (timerfd_settime(fd, TFD_TIMER_ABSTIME|TFD_TIMER_CANCEL_ON_SET, &its32, NULL) >= 0)
return TAKE_FD(fd);
}
#endif
return -errno;
}