Messages in this thread | | | Date | Sun, 30 Jun 2013 18:46:34 +0100 | Subject | Re: [PATCH] proc: make high precision system boot time available | From | Sami Kerola <> |
| |
On 30 June 2013 16:10, Kay Sievers <kay@vrfy.org> wrote: > On Sat, Jun 29, 2013 at 10:53 PM, Sami Kerola <kerolasa@iki.fi> wrote: > >> BTW having a way to measure effect of suspend/resume could lead to a >> way to fix time time distortion. > >> Perhaps there is better alternative to fix user space programs. >> Unfortunately I do not have either knowledge, or imagination, to come >> up with any. Fix hints, ideas, and other proposal are most welcome. > > Stuff should just use: > clock_gettime(CLOCK_BOOTTIME, &ts); > > to get the time in better resolution, and with suspend time taken into account? > > As a general note, the kmsg/printk() time in the raw kernel log is the > time of the CPU it runs on, especially during early init this time > might not be in sync across CPUs and stuff jumps event in the very > same stream on events from the kernel itself: > http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/kernel/sched/clock.c#n329
Thank you Kay,
The clock_gettime(CLOCK_BOOTTIME, &uptime); seems to be exactly what dmesg needs. A patch to fix dmesg wigging will be sent to util-linux maillist shortly. Perhaps the dmesg / journald time stamp difference issue will be fixed someday, but that discussion does not belong to linux-kernel mailist, so it is time to declare
end of thread.
-- Sami Kerola http://www.iki.fi/kerolasa/
| |