[Devel] Re: Q: How complete is the pid namespace in mainline

[Eric W. Biederman]

Cedric Le Goater <clg at fr.ibm.com> writes:

> Eric W. Biederman wrote:
>> Guys how complete do you fee the pid namespace support is that
>> has been merged into Linus's tree?
>> 
>> My impression until I started reading through code earlier today
>> was that the support was just about done except for a couple of
>> tricky details.
>
> Yes It looks sane.
>
> Here's what I have in mind :
>
> * there are some patches from suka that make sure the pid namespace init 
>   is not getting abusively killed by one of this children 
> * the pid cleanup is not complete 
> 	. locks
> 	. kthread (i should work soon on improving kthread to support 
>           signals)
>
> IMO, the proc mount shouldn't be under the pid namespace. we will 
> need that sooner or later.

I was hoping to get a larger list of unfixed issues.

Currently from my review I have generated about 25 bugfix patches.
Several of them in some fairly obvious places.

I think it is a good base to build on, but it feels to like we still
have a significant ways to go.

I think the assumption that we can use global pid numbers instead
of instead of struct pids is racy, and a serious maintenance problem.
It leads to silent breakage of routines like get_net_ns_by_pid,
and possibly a couple of other places.

I'm really not happy with pid_nr meaning a pid number in the
init_pid_ns and pid_vnr meaning a pid number meaning a pid in the
local pid namespace.  But that is just a matter of names so I
don't think it has caused any problems.   Short of making it
to easy to get a pid number in the &init_pid_ns. 

Eric
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers