Re: + prctl-add-pr_setget_child_reaper-to-allow-simple-process-supervision .patch added to -mm tree

[Kay Sievers <kay.sievers@vrfy.org>]

On Thu, Aug 18, 2011 at 23:23, Linus Torvalds <torvalds@linux-foundation.org> wrote:
> On Thu, Aug 18, 2011 at 11:11 AM, Kay Sievers <kay.sievers@vrfy.org> wrote:
>>
>> How about this? It:
>> - uses task->real_parent to walk up the chain of parents.
>
> If I read this right, it does that for all the normal cases too.
> Disgusting. Slowing down the *usual* UNIX case for your new made-up
> case that nobody actually uses is not acceptable.

Re-parenting is not a *usual* operation, usual exit()s have a parent and
do not trigger that code. And in most cases of double-forking it is a
one step, which is almost the same than it was before. Only if we skip
multiple parents it will get the check, which is cheap anyway.

> So NAK. We're not slowing down normal code for some new feature that
> realistically will never be used by any normal applications.

It will be used for all new service managers. UNIX is a pain if you want
to watch your children regarding double-forking, if you are not init.

And it will clean up the mess the desktop stuff is putting in 'ps afx'.
Here is the before/after output:

  253 ?        Ss     0:00 /bin/dbus-daemon --system --address=systemd: --nofork --systemd-activation
  294 ?        Sl     0:00 /usr/libexec/polkit-1/polkitd
  328 ?        S      0:00 /usr/sbin/modem-manager
  608 ?        Sl     0:00 /usr/libexec/colord
  658 ?        Sl     0:00 /usr/libexec/upowerd
  819 ?        Sl     0:00 /usr/libexec/imsettings-daemon
  916 ?        Sl     0:00 /usr/libexec/udisks-daemon
  917 ?        S      0:00  \_ udisks-daemon: not polling any devices


  294 ?        Ss     0:00 /bin/dbus-daemon --system --address=systemd: --nofork --systemd-activation
  426 ?        Sl     0:00  \_ /usr/libexec/polkit-1/polkitd
  449 ?        S      0:00  \_ /usr/sbin/modem-manager
  635 ?        Sl     0:00  \_ /usr/libexec/colord
  705 ?        Sl     0:00  \_ /usr/libexec/upowerd
  959 ?        Sl     0:00  \_ /usr/libexec/udisks-daemon
  960 ?        S      0:00  |   \_ udisks-daemon: not polling any devices
  977 ?        Sl     0:00  \_ /usr/libexec/packagekitd

It's a pretty nice way for an admin too see what's going on on that box,
even with good old 'ps'.

It will also dramatically simplify the handling of started services from
a service manager which double-fork for only historic reasons and make
proper service supervision a nightmare if you are not PID 1.

Upstart ptraces all started services to accomplish something similar. I
doubt we want to go that road, but if you prefer ... :)

If you want, we can make the forked processes inherit a flag if the
'subreaper' should be looked up at all. Then we have almost zero
overhead if the feature isn't used. You think that's needed?

Thanks,
Kay