Re: [PATCH 1/7] cpu-timers: high-resolution CPU clocks for POSIX clock_* syscalls

[Christoph Lameter <clameter@sgi.com>]

On Tue, 14 Dec 2004, Roland McGrath wrote:

> I believe Christoph may have been referring exclusively to the per-process
> clocks, not the per-thread clocks.  I'm not entirely sure since he used the
> term "process" exclusively, but quoted my paragraph about the per-thread
> clock access.  Ulrich's reply is apropos to the individual thread clocks.
>
> It was about the per-process clocks that I raised the question.  For those,
> POSIX says it's implementation-defined what process can see the CPU clock
> of another process.  That means we can make it as restricted or as free as
> we like, but we are obliged to document up front for the users what the
> semantics are.  That's why I would like to make sure we have thought a
> little about the choices now, rather than someone coming along later and
> deciding we really ought to impose security on it (which might be be
> changing the story after developers reasonably coded to the
> implementation-defined behavior we documented in the first place).

The threads are processes under Linux and thus may also be identified by a
pid. You are right, this is a Linux centric view and this identificaition
is not applicable to other systems.