Get the precise time of system bootup on iOS/OS X

Question

Is there an API to obtain the NSDate or NSTimeInterval representing the time the system booted? Some APIs such as [NSProcessInfo systemUptime] and Core Motion return time since boot. I need to precisely correlate these uptime values with NSDates, to about a millisecond.

Time since boot ostensibly provides more precision, but it's easy to see that NSDate already provides precision on the order of 100 nanoseconds, and anything under a microsecond is just measuring interrupt latency and PCB clock jitter.

The obvious thing is to subtract the uptime from the current time [NSDate date]. But that assumes that time does not change between the two system calls, which is, well, hard to accomplish. Moreover if the thread is preempted between the calls, everything is thrown off. The workaround is to repeat the process several times and use the smallest result, but yuck.

NSDate must have a master offset it uses to generate objects with the current time from the system uptime, is there really no way to obtain it?

Answer 1

In OSX you could use sysctl(). This is how the OSX Unix utility uptime does it. Source code is available - search for boottime.

Fair warning though, in iOS i have no idea if this would work.

UPDATE: found some code :)

#include <sys/types.h>
#include <sys/sysctl.h>  

#define MIB_SIZE 2  

int mib[MIB_SIZE];
size_t size;
struct timeval  boottime;

mib[0] = CTL_KERN;
mib[1] = KERN_BOOTTIME;
size = sizeof(boottime);
if (sysctl(mib, MIB_SIZE, &boottime, &size, NULL, 0) != -1)
{
    // successful call
    NSDate* bootDate = [NSDate dateWithTimeIntervalSince1970:
                               boottime.tv_sec + boottime.tv_usec / 1.e6];
}

see if this works...

Answer 2

The accepted answer, using systcl, works, but the values returned by sysctl for KERN_BOOTTIME, at least in my testing (Darwin Kernel Version 11.4.2), are always in whole seconds (the microseconds field, tv_usec, is 0). This means the resulting time may be up to 1 second off, which is not very accurate.

Also, having compared that value, to one derived experimentally from the difference between the REALTIME_CLOCK and CALENDAR_CLOCK, they sometimes differ by a couple seconds, so its not clear whether the KERN_BOOTTIME value corresponds exactly to the time-basis for the uptime clocks.

Answer 3

There is another way. It could give result slightly different (less or more) than accepted answer

I have compared them. I get difference -7 second for OSX 10.9.3 and +2 second for iOS 7.1.1

As i understand this way gives same result if wall clock changed, but accepted answer gives different results if wall clock changed...

Here code:

static CFAbsoluteTime getKernelTaskStartTime(void) {
    enum { MICROSECONDS_IN_SEC = 1000 * 1000 };
    struct kinfo_proc   info;
    bzero(&info, sizeof(info));

    // Initialize mib, which tells sysctl the info we want, in this case
    // we're looking for information about a specific process ID = 0.
    int mib[] = {CTL_KERN, KERN_PROC, KERN_PROC_PID, 0};

    // Call sysctl.
    size_t size = sizeof(info);
    const int sysctlResult = sysctl(mib, COUNT_ARRAY_ELEMS(mib), &info, &size, NULL, 0);
    assert(0 != sysctlResult);

    const struct timeval * timeVal = &(info.kp_proc.p_starttime);
    NSTimeInterval result = -kCFAbsoluteTimeIntervalSince1970;
    result += timeVal->tv_sec;
    result += timeVal->tv_usec / (double)MICROSECONDS_IN_SEC;
    return result;
}

Answer 4

Refer to this category

NSDate+BootTime.h

#import <Foundation/Foundation.h>

@interface NSDate (BootTime)

+ (NSDate *)bootTime;

+ (NSTimeInterval)bootTimeTimeIntervalSinceReferenceDate;

@end

NSDate+BootTime.m

#import "NSDate+BootTime.h"

#include <sys/types.h>
#include <sys/sysctl.h>

@implementation NSDate (BootTime)

+ (NSDate *)bootTime {
    return [NSDate dateWithTimeIntervalSinceReferenceDate:[NSDate bootTimeTimeIntervalSinceReferenceDate]];
}

+ (NSTimeInterval)bootTimeTimeIntervalSinceReferenceDate {
    return getKernelTaskStartTime();
}

////////////////////////////////////////////////////////////////////////
#pragma mark - Private
////////////////////////////////////////////////////////////////////////

#define COUNT_ARRAY_ELEMS(arr) sizeof(arr)/sizeof(arr[0])

static CFAbsoluteTime getKernelTaskStartTime(void) {
    enum { MICROSECONDS_IN_SEC = 1000 * 1000 };
    struct kinfo_proc   info;
    bzero(&info, sizeof(info));

    // Initialize mib, which tells sysctl the info we want, in this case
    // we're looking for information about a specific process ID = 0.
    int mib[] = {CTL_KERN, KERN_PROC, KERN_PROC_PID, 0};

    // Call sysctl.
    size_t size = sizeof(info);
    const int sysctlResult = sysctl(mib, COUNT_ARRAY_ELEMS(mib), &info, &size, NULL, 0);
    if (sysctlResult != -1) {

        const struct timeval * timeVal = &(info.kp_proc.p_starttime);
        NSTimeInterval result = -kCFAbsoluteTimeIntervalSince1970;
        result += timeVal->tv_sec;
        result += timeVal->tv_usec / (double)MICROSECONDS_IN_SEC;
        return result;

    }

    return 0;
}

@end

Answer 5

The routines inside mach/mach_time.h are guaranteed to be monotonically increasing, unlike NSDate.