c/r, time, and virtualization

c/r, time, and virtualization

[Oren Laadan]

Hi,

I wonder what's that status of time namespaces and virtual time(rs) ?

http://lwn.net/Articles/180375/  - "virtual time" coverage in LWN (2006)
http://lkml.org/lkml/2006/4/13/172  - patchset and thread on LKML
http://lwn.net/Articles/249080/  - kernel summit 2007 - containers

It will be handy for c/r of timers (posix-timers) that I'm working on.

A couple of obstacles off the top of my head:

- Lack of clone flags -- we can use clone_with_pids() for that.

- Fast gettimeofday() via vdso -- perhaps a vdso page per time-ns
(containing the necessary deltas).

Is anyone working on it ?

Oren.

Re: c/r, time, and virtualization

[Eric W. Biederman]

Oren Laadan <orenl-RdfvBDnrOixBDgjK7y7TUQ@public.gmane.org> writes:

> Hi,
>
> I wonder what's that status of time namespaces and virtual time(rs) ?
>
> http://lwn.net/Articles/180375/  - "virtual time" coverage in LWN (2006)
> http://lkml.org/lkml/2006/4/13/172  - patchset and thread on LKML
> http://lwn.net/Articles/249080/  - kernel summit 2007 - containers
>
> It will be handy for c/r of timers (posix-timers) that I'm working on.
>
> A couple of obstacles off the top of my head:
>
> - Lack of clone flags -- we can use clone_with_pids() for that.
>
> - Fast gettimeofday() via vdso -- perhaps a vdso page per time-ns
> (containing the necessary deltas).
>
> Is anyone working on it ?

I proposed it, for the reason you care.
I have thought about the semantics.
But I haven't gotten as far as implementing something.

I will be happy to bounce ideas off, and review code
when I have a few seconds free.

Eric