Re: RFC [patch 13/34] PID Virtualization Define new task_pid api

[Dave Hansen <haveblue@us.ibm.com>]

On Tue, 2006-01-17 at 16:03 +0000, Alan Cox wrote: 
> On Maw, 2006-01-17 at 09:56 -0600, Serge E. Hallyn wrote:
> > The virtual pid is different depending on who is asking.  So simply
> > storing current->realpid and current->pid isn't helpful, as we would
> > still need to call a function when a pid crosses user->kernel boundary.
> 
> This is an obscure, weird piece of functionality for some special case
> usages most of which are going to be eliminated by Xen. I don't see the
> kernel side justification for it at all.

At least OpenVZ and vserver want very similar functionality.  They're
both working with out-of-tree patch sets.  We each want to do subtly
different things with tsk->pid, and task_pid() seemed to be a decent
place to start.  OpenVZ has a very similar concept in its pid_task()
function.

Arjan had a very good point last time we posted these: we should
consider getting rid of as many places in the kernel where pids are used
to uniquely identify tasks, and just stick with task_struct pointers.  

> Maybe you should remap it the other side of the user->kernel boundary ?

We would very much like to run unmodified applications and libraries.
Doing it in a patched libc is certainly a possibility, but it wouldn't
make any future customers very happy.

I also wonder if RedHat or SUSE would ever ship and support a special
set of libraries for us.  Oh, and there are always statically linked
apps... :)

-- Dave

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/