On Mon, 5 Mar 2012 14:17:07 -0700, Mark Post wrote: > >Most Linux operating systems read the hardware clock during the startup >process, and use that to set the system clock. NTP takes over from there, but >only affects the system clock, not the hardware clock. The system only tries >to set the hardware clock when shutting down. > But that "system clock" needs some sort of frequency reference. I wouldn't imagine that polling NTP provides fine enough granularity. What does it use?
>There is only one source of time on Linux, and that's from the kernel via the >gettimeofday syscall. So, all processes go to the same place to get their >time information. > What a concept! You mean none of them do STCK(E) and variously apply corrections from the CVT, sometimes correctly; sometimes not? What if they're running in SRB or whatever mode, and can't issue SVC? (Never mind!) >There were some changes to the kernel a while ago that made it aware of >ETR/STP. I never looked at the code to see if it involved any attempts at >modifying the hardware clock. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
