Re: [PATCH v4] pidns: introduce syscall translate_pid

[Konstantin Khlebnikov]

On 17.10.2017 00:05, Nagarathnam Muthusamy wrote:
> On 10/16/2017 09:24 AM, Oleg Nesterov wrote:
> > On 10/13, Konstantin Khlebnikov wrote:
> > > On 13.10.2017 19:05, Oleg Nesterov wrote:
> > > > I won't insist, but this suggests we should add a new helper,
> > > > get_ns_by_fd_type(fd, type), and convert get_net_ns_by_fd() to use it
> > > > as well.
> > > That was in v3.
> > >
> > > I'll prefer to this later, separately. And replace fget with fdget which
> > > allows to do this without atomic operations if task is single-threaded.
> > OK, agreed,
> >
> > > > Stupid question. Can't we make a simpler API which doesn't need /proc/ ?
> > > > I mean,
> > > >
> > > >     sys_translate_pid(pid_t pid, pid_t source_pid, pid_t target_pid)
> > > >     {
> > > >         struct pid_namespace *source_ns, *target_ns;
> > > >
> > > >         source_ns = task_active_pid_ns(find_task_by_vpid(source_pid));
> > > >         target_ns = task_active_pid_ns(find_task_by_vpid(target_pid));
> > > >
> > > >         ...
> > > >     }
> > > > > Yes, this is more limited... Do you have a use-case when this is not enough?
> > > That was in v1 but considered too racy.
> > Hmm, I don't understand...
> >
> > Yes sure, this is racy but open("/proc/$pid/ns/pid") is racy too?
> >
> > OK, once you do fd=open("/proc/$pid/ns/pid") you can use this fd even after
> > its owner exits, while find_task_by_vpid() will fail or find another task if
> > this pid was already reused.
> >
> > But once again, do you have a use-case when this is important?
>
> I believe that in V1 Eric pointed out that pid in general is not a clean way to represent
> namespace. (https://lkml.org/lkml/2015/9/22/1087) Few old interfaces used pid only because at that time there was no better way to represent 
> namespaces.

Yeah, that was a reason.

If we think further - all syscalls who operates with non-child tasks racy and
must be be replaced with some kind of pidfd or taskfd.

Eric pointed that too: https://lkml.org/lkml/2015/9/28/508

> > > But we could merge both ways:
> > >
> > > source >= 0 - pidns fs
> > > source < 0  - task_pid = -source
> > But for what? I must have missed something...

I mean we could have both ways to point namespace in one agrument.
Some classic syscalls emply similar magic for negative pids.

This is cheap and looks almost sane. =)

> > Oleg.