RFC: systemd integration

RFC: systemd integration

[Radu Moisan]

[-- Attachment #1: Type: text/plain, Size: 618 bytes --]

Hi guys,
     I have prepared a branch on poky-contrib (rmoisan/systemd) where I 
pulled systemd from meta-openembedded. The goal is to merge this into 
oe-core, so please take a look and give me your feedback on stuff that 
needs work or things that I might have done wrong, so that I can prepare 
it for merge.
     At this point core-image-minimal and core-image-sato both boot fine 
with systemd. There still are a few warnings but not critical(at 
runtime), at least in my opinion, you may want to take a look at those too.
     You can activate systemd by *DISTRO_FEATURES_INITMAN = "systemd"*

Radu

[-- Attachment #2: Type: text/html, Size: 938 bytes --]

Re: RFC: systemd integration

[Otavio Salvador]

On Wed, Dec 19, 2012 at 3:20 PM, Radu Moisan <radu.moisan@intel.com> wrote:
> Hi guys,
>     I have prepared a branch on poky-contrib (rmoisan/systemd) where I
> pulled systemd from meta-openembedded. The goal is to merge this into
> oe-core, so please take a look and give me your feedback on stuff that needs
> work or things that I might have done wrong, so that I can prepare it for
> merge.
>     At this point core-image-minimal and core-image-sato both boot fine with
> systemd. There still are a few warnings but not critical(at runtime), at
> least in my opinion, you may want to take a look at those too.
>     You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"

In this case, systemd is made *available* or used? Is it possible to
have sysvinit image and another systemd?

--
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: RFC: systemd integration

[Eric Bénard]

Le Wed, 19 Dec 2012 15:21:14 -0200,
Otavio Salvador <otavio@ossystems.com.br> a écrit :

> On Wed, Dec 19, 2012 at 3:20 PM, Radu Moisan <radu.moisan@intel.com> wrote:
> > Hi guys,
> >     I have prepared a branch on poky-contrib (rmoisan/systemd) where I
> > pulled systemd from meta-openembedded. The goal is to merge this into
> > oe-core, so please take a look and give me your feedback on stuff that needs
> > work or things that I might have done wrong, so that I can prepare it for
> > merge.
> >     At this point core-image-minimal and core-image-sato both boot fine with
> > systemd. There still are a few warnings but not critical(at runtime), at
> > least in my opinion, you may want to take a look at those too.
> >     You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"
> 
> In this case, systemd is made *available* or used? Is it possible to
> have sysvinit image and another systemd?
> 
there is a DISTRO_FEATURE in this patch which seems to give this
choice :
http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=rmoisan/systemd&id=1dd15bf46003d0563e7f3441a6c73cbc1e90a79b
DISTRO_FEATURES_INITMAN ?= "sysvinit"

Eric

Re: RFC: systemd integration

[Eric Bénard]

Hi,

Le Wed, 19 Dec 2012 19:20:45 +0200,
Radu Moisan <radu.moisan@intel.com> a écrit :
>      I have prepared a branch on poky-contrib (rmoisan/systemd) where I 
> pulled systemd from meta-openembedded. The goal is to merge this into 
> oe-core, so please take a look and give me your feedback on stuff that 
> needs work or things that I might have done wrong, so that I can prepare 
> it for merge.

a nitpick :
http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=rmoisan/systemd&id=02f4027f7ccddaff3b0316962e7a86a04d31abbc
here you add that to lighttpd recipe :
+ file://lighttpd_fix_for_automake-1.12.patch \
+ file://lighttpd.service
but you don't add the patch

then here (several patches later)
http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=rmoisan/systemd&id=0b6a5d3f998d2f1bb346c403d512c33ee93052fb
you fix that :
- file://lighttpd_fix_for_automake-1.12.patch \


>      At this point core-image-minimal and core-image-sato both boot fine 
> with systemd. There still are a few warnings but not critical(at 
> runtime), at least in my opinion, you may want to take a look at those too.
>      You can activate systemd by *DISTRO_FEATURES_INITMAN = "systemd"*
> 
thanks, will try this.

Eric

Re: RFC: systemd integration

[Radu Moisan]

On 12/19/2012 10:57 PM, Eric Bénard wrote:
> Hi,
>
> Le Wed, 19 Dec 2012 19:20:45 +0200,
> Radu Moisan <radu.moisan@intel.com> a écrit :
>>       I have prepared a branch on poky-contrib (rmoisan/systemd) where I
>> pulled systemd from meta-openembedded. The goal is to merge this into
>> oe-core, so please take a look and give me your feedback on stuff that
>> needs work or things that I might have done wrong, so that I can prepare
>> it for merge.
> a nitpick :
> http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=rmoisan/systemd&id=02f4027f7ccddaff3b0316962e7a86a04d31abbc
> here you add that to lighttpd recipe :
> + file://lighttpd_fix_for_automake-1.12.patch \
> + file://lighttpd.service
> but you don't add the patch
>
> then here (several patches later)
> http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=rmoisan/systemd&id=0b6a5d3f998d2f1bb346c403d512c33ee93052fb
> you fix that :
> - file://lighttpd_fix_for_automake-1.12.patch \
that was a regression I introduced when I rebased on master a while 
back. I'll squash it with the initial commit.

Radu

Re: RFC: systemd integration

[Radu Moisan]

On 12/19/2012 07:21 PM, Otavio Salvador wrote:
> On Wed, Dec 19, 2012 at 3:20 PM, Radu Moisan <radu.moisan@intel.com> wrote:
>> Hi guys,
>>      I have prepared a branch on poky-contrib (rmoisan/systemd) where I
>> pulled systemd from meta-openembedded. The goal is to merge this into
>> oe-core, so please take a look and give me your feedback on stuff that needs
>> work or things that I might have done wrong, so that I can prepare it for
>> merge.
>>      At this point core-image-minimal and core-image-sato both boot fine with
>> systemd. There still are a few warnings but not critical(at runtime), at
>> least in my opinion, you may want to take a look at those too.
>>      You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"
> In this case, systemd is made *available* or used? Is it possible to
> have sysvinit image and another systemd?
it is used.

Radu

Re: RFC: systemd integration

[Yi Qingliang]

[-- Attachment #1: Type: text/plain, Size: 787 bytes --]

GREAT!!!

I have asked systemd status in yocto, maybe I can use your branch instead!

THANKS!


On Wednesday, December 19, 2012 07:20:45 PM Radu Moisan wrote:

Hi guys,
    I have prepared a branch on poky-contrib (rmoisan/systemd) where I pulled 
systemd from meta-openembedded. The goal is to merge this into oe-core, so 
please take a look and give me your feedback on stuff that needs work or things 
that I might have done wrong, so that I can prepare it for merge.
    At this point core-image-minimal and core-image-sato both boot fine with 
systemd. There still are a few warnings but not critical(at runtime), at least 
in my opinion, you may want to take a look at those too.
    You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"

Radu




[-- Attachment #2: Type: text/html, Size: 2688 bytes --]

Re: RFC: systemd integration

[Radu Moisan]

On 12/20/2012 10:52 AM, Radu Moisan wrote:
>
> On 12/19/2012 10:57 PM, Eric Bénard wrote:
>> Hi,
>>
>> Le Wed, 19 Dec 2012 19:20:45 +0200,
>> Radu Moisan <radu.moisan@intel.com> a écrit :
>>>       I have prepared a branch on poky-contrib (rmoisan/systemd) 
>>> where I
>>> pulled systemd from meta-openembedded. The goal is to merge this into
>>> oe-core, so please take a look and give me your feedback on stuff that
>>> needs work or things that I might have done wrong, so that I can 
>>> prepare
>>> it for merge.
>> a nitpick :
>> http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=rmoisan/systemd&id=02f4027f7ccddaff3b0316962e7a86a04d31abbc 
>>
>> here you add that to lighttpd recipe :
>> + file://lighttpd_fix_for_automake-1.12.patch \
>> + file://lighttpd.service
>> but you don't add the patch
>>
>> then here (several patches later)
>> http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/?h=rmoisan/systemd&id=0b6a5d3f998d2f1bb346c403d512c33ee93052fb 
>>
>> you fix that :
>> - file://lighttpd_fix_for_automake-1.12.patch \
> that was a regression I introduced when I rebased on master a while 
> back. I'll squash it with the initial commit.
     It just got clearer what happened here. So, there was a regression, 
systemd related changes were missing after rebase. However with automake 
patch something alse happened.  Since my initial commit, lighttpd was 
updated to 1.4.32 and it didn't need that patch anymore, so it got 
removed. When I rebased, the patch was removed but the inclusion in .bb 
file didn't so I got an error, which I had to fix.
     For consistency, I'll squash systemd related changes back into the 
first commit, but the removal of automake patch will stay in this last 
patch.

> Radu
>
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

Re: RFC: systemd integration

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 1802 bytes --]

On Wed, Dec 19, 2012 at 07:20:45PM +0200, Radu Moisan wrote:
> Hi guys,
>      I have prepared a branch on poky-contrib (rmoisan/systemd) where I 
> pulled systemd from meta-openembedded. The goal is to merge this into 
> oe-core, so please take a look and give me your feedback on stuff that 
> needs work or things that I might have done wrong, so that I can prepare 
> it for merge.
>      At this point core-image-minimal and core-image-sato both boot fine 
> with systemd. There still are a few warnings but not critical(at 
> runtime), at least in my opinion, you may want to take a look at those too.
>      You can activate systemd by *DISTRO_FEATURES_INITMAN = "systemd"*

What about upgrade path from meta-systemd?

I've noticed that
http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/meta/recipes-connectivity/wpa-supplicant?h=rmoisan/systemd&id=02f4027f7ccddaff3b0316962e7a86a04d31abbc
does not increment PR at all while
http://git.openembedded.org/meta-openembedded/tree/meta-systemd/oe-core/recipes-connectivity/wpa-supplicant/wpa-supplicant_1.0.bbappend
PRINC := "${@int(PRINC) + 2}"

Also I like this commit very much
http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/meta/recipes-connectivity/wpa-supplicant?h=rmoisan/systemd&id=216049b154799fc7984c54858a386d7318f99ce0
but current styleguide thinks something else about consistent whitespace
usage. So I think this wont be accepted as my change to styleguide
wasn't.

And last thing:
there is few pending patches for meta-systemd:
http://patchwork.openembedded.org/project/oe/list/?q=systemd
Can you check them and merge them if needed? Or will you update your
systemd* copies if we merge them in meta-oe?

Cheers,

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 205 bytes --]

Re: RFC: systemd integration

[Koen Kooi]

Op 19 dec. 2012, om 18:20 heeft Radu Moisan <radu.moisan@intel.com> het volgende geschreven:

> Hi guys,
>     I have prepared a branch on poky-contrib (rmoisan/systemd) where I pulled systemd from meta-openembedded. The goal is to merge this into oe-core, so please take a look and give me your feedback on stuff that needs work or things that I might have done wrong, so that I can prepare it for merge.
>     At this point core-image-minimal and core-image-sato both boot fine with systemd. There still are a few warnings but not critical(at runtime), at least in my opinion, you may want to take a look at those too.
>     You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"

What happens when I do DISTRO_FEATURES_INITMAN = "sysvinit systemd"?

Re: RFC: systemd integration

[Radu Moisan]

On 12/20/2012 12:47 PM, Koen Kooi wrote:
> Op 19 dec. 2012, om 18:20 heeft Radu Moisan <radu.moisan@intel.com> het volgende geschreven:
>
>> Hi guys,
>>      I have prepared a branch on poky-contrib (rmoisan/systemd) where I pulled systemd from meta-openembedded. The goal is to merge this into oe-core, so please take a look and give me your feedback on stuff that needs work or things that I might have done wrong, so that I can prepare it for merge.
>>      At this point core-image-minimal and core-image-sato both boot fine with systemd. There still are a few warnings but not critical(at runtime), at least in my opinion, you may want to take a look at those too.
>>      You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"
> What happens when I do DISTRO_FEATURES_INITMAN = "sysvinit systemd"?
I think it would break your build. But what would be the point of doing 
that?

Radu

Re: RFC: systemd integration

[Radu Moisan]

On 12/20/2012 12:42 PM, Martin Jansa wrote:
> On Wed, Dec 19, 2012 at 07:20:45PM +0200, Radu Moisan wrote:
>> Hi guys,
>>       I have prepared a branch on poky-contrib (rmoisan/systemd) where I
>> pulled systemd from meta-openembedded. The goal is to merge this into
>> oe-core, so please take a look and give me your feedback on stuff that
>> needs work or things that I might have done wrong, so that I can prepare
>> it for merge.
>>       At this point core-image-minimal and core-image-sato both boot fine
>> with systemd. There still are a few warnings but not critical(at
>> runtime), at least in my opinion, you may want to take a look at those too.
>>       You can activate systemd by *DISTRO_FEATURES_INITMAN = "systemd"*
> What about upgrade path from meta-systemd?
>
> I've noticed that
> http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/meta/recipes-connectivity/wpa-supplicant?h=rmoisan/systemd&id=02f4027f7ccddaff3b0316962e7a86a04d31abbc
> does not increment PR at all while
> http://git.openembedded.org/meta-openembedded/tree/meta-systemd/oe-core/recipes-connectivity/wpa-supplicant/wpa-supplicant_1.0.bbappend
> PRINC := "${@int(PRINC) + 2}"
I'll take care of this one.
>
> Also I like this commit very much
> http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/meta/recipes-connectivity/wpa-supplicant?h=rmoisan/systemd&id=216049b154799fc7984c54858a386d7318f99ce0
> but current styleguide thinks something else about consistent whitespace
> usage. So I think this wont be accepted as my change to styleguide
> wasn't.
I think Richard suggested it, but anyway wasn't the recomandation to use 
spaces instead of tabs (I use 4 spaces)? I can revert or change it when 
time comes. Feedback on this will be well appreciated.
> And last thing:
> there is few pending patches for meta-systemd:
> http://patchwork.openembedded.org/project/oe/list/?q=systemd
> Can you check them and merge them if needed? Or will you update your
> systemd* copies if we merge them in meta-oe?
These look very interesting. I'll take a closer look and pull them in as 
appropriate.

Radu

Re: RFC: systemd integration

[Burton, Ross]

On 20 December 2012 10:47, Koen Kooi <koen@dominion.thruhere.net> wrote:
>>     You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"
>
> What happens when I do DISTRO_FEATURES_INITMAN = "sysvinit systemd"?

As systemd is showing, the choice of init manager isn't just something
you can do at image generation time, but impacts the build.  Services
may opt to use systemd-native interfaces if systemd is available, or
streamline their daemonisation based on the support that systemd has.

This is why this series makes the choice of init a distribution
choice, and doesn't allow multiple choices.

Ross

Re: RFC: systemd integration

[Otavio Salvador]

On Thu, Dec 20, 2012 at 10:39 AM, Burton, Ross <ross.burton@intel.com> wrote:
> On 20 December 2012 10:47, Koen Kooi <koen@dominion.thruhere.net> wrote:
>>>     You can activate systemd by  DISTRO_FEATURES_INITMAN = "systemd"
>>
>> What happens when I do DISTRO_FEATURES_INITMAN = "sysvinit systemd"?
>
> As systemd is showing, the choice of init manager isn't just something
> you can do at image generation time, but impacts the build.  Services
> may opt to use systemd-native interfaces if systemd is available, or
> streamline their daemonisation based on the support that systemd has.
>
> This is why this series makes the choice of init a distribution
> choice, and doesn't allow multiple choices.

Not really, I've been using sysvinit images and systemd images in OE
with meta-oe and it works fine.

There're some changes need, indeed, but it is possible to make it
work. The biggest change I had to do is to get systemd-udevd using
init scripts.

--
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: RFC: systemd integration

[Burton, Ross]

On 20 December 2012 12:50, Otavio Salvador <otavio@ossystems.com.br> wrote:
> Not really, I've been using sysvinit images and systemd images in OE
> with meta-oe and it works fine.
>
> There're some changes need, indeed, but it is possible to make it
> work. The biggest change I had to do is to get systemd-udevd using
> init scripts.

Sure it works, I never said it didn't work. It's just not optimal -
dbus isn't using the superior console ownership information available
to it through systemd for example, because that means linking to
libsystemd-login.  By making systemd an image-time choice we can't
ever have any code that requires systemd in the build.

Ross

Re: RFC: systemd integration

[Otavio Salvador]

On Thu, Dec 20, 2012 at 10:59 AM, Burton, Ross <ross.burton@intel.com> wrote:
> On 20 December 2012 12:50, Otavio Salvador <otavio@ossystems.com.br> wrote:
>> Not really, I've been using sysvinit images and systemd images in OE
>> with meta-oe and it works fine.
>>
>> There're some changes need, indeed, but it is possible to make it
>> work. The biggest change I had to do is to get systemd-udevd using
>> init scripts.
>
> Sure it works, I never said it didn't work. It's just not optimal -
> dbus isn't using the superior console ownership information available
> to it through systemd for example, because that means linking to
> libsystemd-login.  By making systemd an image-time choice we can't
> ever have any code that requires systemd in the build.

I think if I enable systemd feature I do expect to have some systemd
libraries in rootfs however I may opt to use it or not. If I do want
to have *no* systemd libraries, than I should not have systemd feature
enabled.

--
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: RFC: systemd integration

[Burton, Ross]

On 20 December 2012 13:09, Otavio Salvador <otavio@ossystems.com.br> wrote:
> I think if I enable systemd feature I do expect to have some systemd
> libraries in rootfs however I may opt to use it or not. If I do want
> to have *no* systemd libraries, than I should not have systemd feature
> enabled.

What's the point of enabling systemd, having the libraries installed,
and then not using it?

I don't see what the appeal of supporting multiple init systems a
single distribution, with the downsides of doubling the test space and
wasting space in situations where you ship all the systemd libraries
but don't actually boot with systemd.

Ross

Re: RFC: systemd integration

[Otavio Salvador]

On Thu, Dec 20, 2012 at 12:31 PM, Burton, Ross <ross.burton@intel.com> wrote:
> On 20 December 2012 13:09, Otavio Salvador <otavio@ossystems.com.br> wrote:
>> I think if I enable systemd feature I do expect to have some systemd
>> libraries in rootfs however I may opt to use it or not. If I do want
>> to have *no* systemd libraries, than I should not have systemd feature
>> enabled.
>
> What's the point of enabling systemd, having the libraries installed,
> and then not using it?
>
> I don't see what the appeal of supporting multiple init systems a
> single distribution, with the downsides of doubling the test space and
> wasting space in situations where you ship all the systemd libraries
> but don't actually boot with systemd.

Here at O.S. Systems we share same distro and packages among different
products and some use sysvinit and other systemd. I think it is a
valid use case.

--
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: RFC: systemd integration

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 1113 bytes --]

On Wed, Dec 19, 2012 at 07:20:45PM +0200, Radu Moisan wrote:
> Hi guys,
>      I have prepared a branch on poky-contrib (rmoisan/systemd) where I 
> pulled systemd from meta-openembedded. The goal is to merge this into 
> oe-core, so please take a look and give me your feedback on stuff that 
> needs work or things that I might have done wrong, so that I can prepare 
> it for merge.
>      At this point core-image-minimal and core-image-sato both boot fine 
> with systemd. There still are a few warnings but not critical(at 
> runtime), at least in my opinion, you may want to take a look at those too.
>      You can activate systemd by *DISTRO_FEATURES_INITMAN = "systemd"*

Is this correct?

http://git.yoctoproject.org/cgit/cgit.cgi/poky-contrib/commit/meta/recipes-extended/pam/libpam_1.1.6.bb?h=rmoisan/systemd&id=d31d804ad76829de039f3b41d7b45ffcc25e69b5

Shouldn't it be added to common-session only when systemd is enabled?

Is '|-' some special syntax for palm config?

"session required|-pam_systemd.so" 

Cheers,
-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 205 bytes --]