[thud] systemd: RDEPENDS on util-linux-umount

Submitted by Adrian Bunk on March 7, 2019, 2:27 p.m. | Patch ID: 159400

Details

Message ID 20190307142709.12076-1-bunk@stusta.de
State Thud Next
Delegated to: Armin Kuster
Headers show

Commit Message

Adrian Bunk March 7, 2019, 2:27 p.m.
From: André Draszik <andre.draszik@jci.com>

It looks like there is an implicit dependency on util-linux'
umount - as otherwise when using busybox' umount we see a
long delay on shutdown / reboot.

[YOCTO #13058]

(From OE-Core rev: 39a3d2c603429865af632fe41b2cf32c3dfdfb1d)

Signed-off-by: André Draszik <andre.draszik@jci.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
---
 meta/recipes-core/systemd/systemd_239.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Patch hide | download patch | download mbox

diff --git a/meta/recipes-core/systemd/systemd_239.bb b/meta/recipes-core/systemd/systemd_239.bb
index ed10f634b9..edda923a34 100644
--- a/meta/recipes-core/systemd/systemd_239.bb
+++ b/meta/recipes-core/systemd/systemd_239.bb
@@ -544,7 +544,7 @@  FILES_${PN} = " ${base_bindir}/* \
 
 FILES_${PN}-dev += "${base_libdir}/security/*.la ${datadir}/dbus-1/interfaces/ ${sysconfdir}/rpm/macros.systemd"
 
-RDEPENDS_${PN} += "kmod dbus util-linux-mount udev (= ${EXTENDPKGV}) util-linux-agetty util-linux-fsck"
+RDEPENDS_${PN} += "kmod dbus util-linux-mount util-linux-umount udev (= ${EXTENDPKGV}) util-linux-agetty util-linux-fsck"
 RDEPENDS_${PN} += "${@bb.utils.contains('PACKAGECONFIG', 'serial-getty-generator', '', 'systemd-serialgetty', d)}"
 RDEPENDS_${PN} += "volatile-binds update-rc.d systemd-conf"
 

Comments

Leon Woestenberg March 7, 2019, 2:56 p.m.
On Thu, Mar 7, 2019 at 3:27 PM Adrian Bunk <bunk@stusta.de> wrote:
>
> From: André Draszik <andre.draszik@jci.com>
>
> It looks like there is an implicit dependency on util-linux'
> umount - as otherwise when using busybox' umount we see a
> long delay on shutdown / reboot.
>
> [YOCTO #13058]
>
That bug number is wrong, seems only slighty related:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=13058

Following the discussions, I think this had to do with older versions
of busybox not ignoring the '-c' option that systemd passes to umount.

https://github.com/systemd/systemd/issues/7786

So, systemd has a dependency on *either* util-linux-mount *or* a
minimal version of busybox. Do we support minimal version dependencies
in Yocto?

Regards,

Leon.
Adrian Bunk March 7, 2019, 3:18 p.m.
On Thu, Mar 07, 2019 at 03:56:33PM +0100, Leon Woestenberg wrote:
> On Thu, Mar 7, 2019 at 3:27 PM Adrian Bunk <bunk@stusta.de> wrote:
> >
> > From: André Draszik <andre.draszik@jci.com>
> >
> > It looks like there is an implicit dependency on util-linux'
> > umount - as otherwise when using busybox' umount we see a
> > long delay on shutdown / reboot.
> >
> > [YOCTO #13058]
> >
> That bug number is wrong, seems only slighty related:
> https://bugzilla.yoctoproject.org/show_bug.cgi?id=13058
> 
> Following the discussions, I think this had to do with older versions
> of busybox not ignoring the '-c' option that systemd passes to umount.
> 
> https://github.com/systemd/systemd/issues/7786
> 
> So, systemd has a dependency on *either* util-linux-mount *or* a
> minimal version of busybox. Do we support minimal version dependencies
> in Yocto?

Then the patch to master didn't make sense,
unfixed busybox is not a problem here.

If systemd needs util-linux mount/umount in master it should also
depend on them in older releases (both master and thud had systemd 239 
at that point).

If it does not, it should not pull in such dependencies unconditionally
in master.

> Regards,
> 
> Leon.

cu
Adrian
Leon Woestenberg March 7, 2019, 3:36 p.m.
Hi Adrian,

On Thu, Mar 7, 2019 at 4:18 PM Adrian Bunk <bunk@stusta.de> wrote:
>
> On Thu, Mar 07, 2019 at 03:56:33PM +0100, Leon Woestenberg wrote:
> > On Thu, Mar 7, 2019 at 3:27 PM Adrian Bunk <bunk@stusta.de> wrote:
> > >
> > > From: André Draszik <andre.draszik@jci.com>
> > >
> > > It looks like there is an implicit dependency on util-linux'
> > > umount - as otherwise when using busybox' umount we see a
> > > long delay on shutdown / reboot.
> > >
> > > [YOCTO #13058]
> > >
> If systemd needs util-linux mount/umount in master it should also
> depend on them in older releases (both master and thud had systemd 239
> at that point).
>
Agreed. I was just adding info to this e-mail thread. The "looks like
there is an implicit dependency" text was a bit too vague for me. I'ld
like to add the root cause that was found.

The implicit dependency was for a specific option to umount to be supported.

If we have the proper busybox in master, we no longer have a implicit
dependency on util-linux-umount.

(The reason for the util-linux-mount dependency was different, AFAIK.)

Regards,

Leon.
André Draszik March 7, 2019, 3:38 p.m.
On Thu, 2019-03-07 at 17:18 +0200, Adrian Bunk wrote:
> On Thu, Mar 07, 2019 at 03:56:33PM +0100, Leon Woestenberg wrote:
> > On Thu, Mar 7, 2019 at 3:27 PM Adrian Bunk <bunk@stusta.de> wrote:
> > > From: André Draszik <andre.draszik@jci.com>
> > > 
> > > It looks like there is an implicit dependency on util-linux'
> > > umount - as otherwise when using busybox' umount we see a
> > > long delay on shutdown / reboot.
> > > 
> > > [YOCTO #13058]
> > > 
> > That bug number is wrong, seems only slighty related:
> > https://bugzilla.yoctoproject.org/show_bug.cgi?id=13058
> > 
> > Following the discussions, I think this had to do with older versions
> > of busybox not ignoring the '-c' option that systemd passes to umount.
> > 
> > https://github.com/systemd/systemd/issues/7786
> > 
> > So, systemd has a dependency on *either* util-linux-mount *or* a
> > minimal version of busybox. Do we support minimal version dependencies
> > in Yocto?
> 
> Then the patch to master didn't make sense,
> unfixed busybox is not a problem here.

Yeah, I did ask for my patch to be ignored, and there was further discussion
confirming that it isn't needed these days. Unfortunately my patch still
went in.

http://lists.openembedded.org/pipermail/openembedded-core/2019-February/278893.html


> If systemd needs util-linux mount/umount in master it should also
> depend on them in older releases (both master and thud had systemd 239 
> at that point).
> 
> If it does not, it should not pull in such dependencies unconditionally
> in master.

Care to post a patch reverting it on master?

Cheers,
Andre'

> 
> > Regards,
> > 
> > Leon.
> 
> cu
> Adrian
> 
> -- 
> 
>        "Is there not promise of rain?" Ling Tan asked suddenly out
>         of the darkness. There had been need of rain for many days.
>        "Only a promise," Lao Er said.
>                                        Pearl S. Buck - Dragon Seed
>
Richard Purdie March 7, 2019, 9:48 p.m.
On Thu, 2019-03-07 at 15:38 +0000, André Draszik wrote:
> On Thu, 2019-03-07 at 17:18 +0200, Adrian Bunk wrote:
> > On Thu, Mar 07, 2019 at 03:56:33PM +0100, Leon Woestenberg wrote:
> > > On Thu, Mar 7, 2019 at 3:27 PM Adrian Bunk <bunk@stusta.de>
> > > wrote:
> > > > From: André Draszik <andre.draszik@jci.com>
> > > > 
> > > > It looks like there is an implicit dependency on util-linux'
> > > > umount - as otherwise when using busybox' umount we see a
> > > > long delay on shutdown / reboot.
> > > > 
> > > > [YOCTO #13058]
> > > > 
> > > That bug number is wrong, seems only slighty related:
> > > https://bugzilla.yoctoproject.org/show_bug.cgi?id=13058
> > > 
> > > Following the discussions, I think this had to do with older
> > > versions
> > > of busybox not ignoring the '-c' option that systemd passes to
> > > umount.
> > > 
> > > https://github.com/systemd/systemd/issues/7786
> > > 
> > > So, systemd has a dependency on *either* util-linux-mount *or* a
> > > minimal version of busybox. Do we support minimal version
> > > dependencies
> > > in Yocto?
> > 
> > Then the patch to master didn't make sense,
> > unfixed busybox is not a problem here.
> 
> Yeah, I did ask for my patch to be ignored, and there was further
> discussion
> confirming that it isn't needed these days. Unfortunately my patch
> still
> went in.
> 
> http://lists.openembedded.org/pipermail/openembedded-core/2019-February/278893.html
> 
> 
> > If systemd needs util-linux mount/umount in master it should also
> > depend on them in older releases (both master and thud had systemd
> > 239 
> > at that point).
> > 
> > If it does not, it should not pull in such dependencies
> > unconditionally
> > in master.
> 
> Care to post a patch reverting it on master?

Going from memory I think there were some responses saying we should
probably add this patch regardless since we already needed util-linux-
mount. I didn't think it was a huge issue since mount was already there
and it seemed to solve some problems others were having.

I don't have a strong preference either way, we just need a general
consensus about what to do...

Cheers,

Richard
Adrian Bunk March 8, 2019, 10:50 a.m.
On Thu, Mar 07, 2019 at 09:48:02PM +0000, Richard Purdie wrote:
> 
> Going from memory I think there were some responses saying we should
> probably add this patch regardless since we already needed util-linux-
> mount. I didn't think it was a huge issue since mount was already there
> and it seemed to solve some problems others were having.
> 
> I don't have a strong preference either way, we just need a general
> consensus about what to do...

The general problem seems to be that systemd upstream does not care much 
whether busybox versions of commands work with systemd.

This creates a conflict between upstream recommendations and aiming at a 
tiny system when using systemd.

The latter would require a continuous effort from someone at OE/Yocto to 
ensure that the runtime dependencies of systemd are small but working.

> Cheers,
> 
> Richard

cu
Adrian