[Buildroot] [PATCH v3] package/systemd: register NSS plugins in nsswitch.conf

Norbert Lange nolange79 at gmail.com
Sat Jul 4 22:17:51 UTC 2020


Am Sa., 4. Juli 2020 um 23:38 Uhr schrieb Yann E. MORIN
<yann.morin.1998 at free.fr>:
>
> Norbert, All,
>
> On 2020-07-04 22:49 +0200, Norbert Lange spake thusly:
> > Am Sa., 4. Juli 2020 um 21:27 Uhr schrieb Yann E. MORIN
> > <yann.morin.1998 at free.fr>:
> > > No, that is exactly what target-finalize is for. ROOTFS_PRE_CMD_HOOKS
> > > are not here to modify the files, but to fixup the layout is some very
> > > rare corner cases (and the rare corner case I am talking about, was
> > > exactly having systemd run on a read-only filesystem with a working
> > > factory; go check the commit logs for the gory details ;-) ).
> >
> > I meant if you need an order like
> > systemd (fixup nsswitch.conf) -> mdns4 (fixup nsswitch.conf) ->
> > systemd (pickup nsswitch.conf *after all modifications*)
> >
> > then you are out of luck, and need to use ROOTFS_PRE_CMD_HOOKS,
>
> Not even, becasue if two packages define ROOTFS_PRE_CMD_HOOKS, then
> those are always run in alphabetical order (based on the package name).
>
> So it's back to square one: there is no way to define dependency order
> for hooks inside a step.
>
> > or use some weird tricks like in [1], where I re-add PURGE_LOCALES.
> > (I will redo those patches, doing that cleanup in systemd.mk).
>
> Yes, I saw it, and I haven't replied yet because:
>
>   - I don't like that nasty trick, but...
>   - I don't have an alternative solution to offer. Not yet...
>
> So I'd have to think about it yet more...
>
> > Yeah, I see'n that trickery with tmpfiles.d ;)
> > But I guess multiple lists of PRE_CMD / TARGET_FINALIZE HOOKS would ease
> > some pains (like adding _EARLY and _LATE variants).
>
> That would not be very helpful either, because we would tehn want to
> have early hooks to run before some other early hooks. This is an
> endless errand...

Sure, but it would be enough to properly fix a few corner cases.

> And in the end, if as a end-user of Buildroot,there are things you
> really want to be done in a specific order, then there always is the
> possibility to provide a post-build script.

And everything you do might depend on several packages, and be broken in
the future. The more you get into buildroot the more reliable (tested) it gets,
only downside is that you have to clean up the mess first ;)

>
> > Btw, I use a different approach by pre-creating all used directories,
> > and mounting /var/tmp as tmpfs. This allows startup without any errors,
> > intended for when it's not possible to mount the usual overlayfs for RW access.
>
> It was my understanding that, should /var be read-only, systemd would
> automatically mount a tmpfs on it and populate it with its counterpart
> from the factory.

Nope, not to my knowlegde. there is the volatile-root service [1]
which can create you a tmpfs
and either use it as overlayfs over the whole rootfs, or AFAIU just
the rootfs /usr directory/mountpoint
will be used (/etc and /var being empty).
Thing is, this only works if systemd is running in the initramfs, Id
guess in buildroot you build either
the real root *or* the initramfs (v246 will have an option 'initrd'
for that reason)

> Or something along those lines...
>
> > > Thanks for your v4! :-)  I'll handle itm soonish.
> > >
> > > BTW, I have a new version of dbus-broker that is working and clean; I'll
> > > send it later too (test running while I write this email).
> >
> > Great, I hope I'll get my stuff upstream in time for systemd 246,
> > should be able to build a pretty lean systemd rootfs with ~16MB size
> > (~23MB today) then.
>
> I would say that size should not be the driving force ehind those
> changes, because 16MiB is already a lot. If you really are after size,
> then sysv-init, busybox, or even openrc are better alternatives (busybox
> would make for a ~1.5MiB runtime on uClibc, so a minimalist systemd-based
> system is already an order of magnitude bigger than that).

it's a matter of functionality vs size, I went from a busybox rootfs to systemd,
after adding supervisord, network supervision, socketactivation and eudev
and tying all together proved nontrivial.
I still dont like having sysv compatibility, dbus and services for
initrd if I dont use it.

> Rather, I think the "surface attack", unused components or unwanted
> components, is a better approach to sell these changes.

Size is a bad but easy to sell measurement for exactly that
(especially for non-techs).

Regards, Norbert

[1] - https://www.freedesktop.org/software/systemd/man/systemd-volatile-root.service.html


More information about the buildroot mailing list