Patchwork [v2] systemd.bbclass: only restart the service when it is active at systemd_postinst

login
register
mail settings
Submitter Li Zhijian
Date Nov. 5, 2013, 10:14 a.m.
Message ID <5278C502.1080100@cn.fujitsu.com>
Download mbox | patch
Permalink /patch/61149/
State New
Headers show

Comments

Li Zhijian - Nov. 5, 2013, 10:14 a.m.
From: Li Zhijian <lizhijian@cn.fujitsu.com>

If systemd_postinst is installed as a delayed package scriptlet (installed to /etc/rpm-postints/${PN})
which would be executed at the first time of OS startup. And then this script would block when it is
trying to call systemctl to restart service.

It seems that it is not a good timing to restart service at run-postinsts.
I test serval services which has supported systemd(uuidd/avahi-daemon), all of them
will block the startup.

Steps to reproduce:
1) Installed systemd_postinst as a delayed package scriptlet
   i.e. Append Following lines to meta/recipes-core/busybox/busybox.inc
pkg_postinst_${PN}-syslog () {
       if [ "x$D" != "x" ] ; then
               exit 1
       fi
}
2) build a rootfs
$ bitbake core-image-minimal
3) startup this rootfs

OS would block at /etc/rpm-postinsts/busybox-syslog, like follows
[ ***] A start job is running ro Run pending postinsts

This patch can fix the block problem, and make the service is consistent with the original status

Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
---
 meta/classes/systemd.bbclass |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

1.7.1
Li Zhijian - Nov. 19, 2013, 2:26 a.m.
ping...
Ross Burton - Jan. 22, 2014, 5:56 p.m.
Whoops, this fell out of my inbox.   Whilst the is-active check now
means that the service is restarted on upgrades, if the package is
being installed it's now not started if AUTO_ENABLE is set to true.

I think the logic should be "$D unset and (auto enable or is-active)"

Ross

On 19 November 2013 02:26, Li Zhijian <lizhijian@cn.fujitsu.com> wrote:
> ping...
>
> --
> Best regards.
> Li Zhijian
>
>  |-----Original Message-----
>  |From: openembedded-core-bounces@lists.openembedded.org
>  |[mailto:openembedded-core-bounces@lists.openembedded.org] On Behalf Of
>  |Li Zhijian
>  |Sent: Tuesday, November 05, 2013 6:14 PM
>  |To: Patches and discussions about the oe-core layer
>  |Subject: [OE-core] [PATCH v2] systemd.bbclass: only restart the service
>  |when it is active at systemd_postinst
>  |
>  |From: Li Zhijian <lizhijian@cn.fujitsu.com>
>  |
>  |If systemd_postinst is installed as a delayed package scriptlet (installed
>  |to /etc/rpm-postints/${PN})
>  |which would be executed at the first time of OS startup. And then this
> script
>  |would block when it is
>  |trying to call systemctl to restart service.
>  |
>  |It seems that it is not a good timing to restart service at run-postinsts.
>  |I test serval services which has supported systemd(uuidd/avahi-daemon),
>  |all of them
>  |will block the startup.
>  |
>  |Steps to reproduce:
>  |1) Installed systemd_postinst as a delayed package scriptlet
>  |   i.e. Append Following lines to meta/recipes-core/busybox/busybox.inc
>  |pkg_postinst_${PN}-syslog () {
>  |       if [ "x$D" != "x" ] ; then
>  |               exit 1
>  |       fi
>  |}
>  |2) build a rootfs
>  |$ bitbake core-image-minimal
>  |3) startup this rootfs
>  |
>  |OS would block at /etc/rpm-postinsts/busybox-syslog, like follows
>  |[ ***] A start job is running ro Run pending postinsts
>  |
>  |This patch can fix the block problem, and make the service is consistent
>  |with the original status
>  |
>  |Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
>  |---
>  | meta/classes/systemd.bbclass |    3 ++-
>  | 1 files changed, 2 insertions(+), 1 deletions(-)
>  |
>  |diff --git a/meta/classes/systemd.bbclass b/meta/classes/systemd.bbclass
>  |index 3700b2e..1feac06 100644
>  |--- a/meta/classes/systemd.bbclass
>  |+++ b/meta/classes/systemd.bbclass
>  |@@ -32,7 +32,8 @@ fi
>  | if type systemctl >/dev/null 2>/dev/null; then
>  |      systemctl $OPTS ${SYSTEMD_AUTO_ENABLE} ${SYSTEMD_SERVICE}
>  | -    if [ -z "$D" -a "${SYSTEMD_AUTO_ENABLE}" = "enable" ]; then
>  |+     if [ -z "$D" ] && systemctl is-active ${SYSTEMD_SERVICE}
>  |&>/dev/null; then
>  |+             # if service is active, restart it
>  |              systemctl restart ${SYSTEMD_SERVICE}
>  |      fi
>  | fi
>  |--
>  |1.7.1
>  |
>  |
>  |
>  |
>  |_______________________________________________
>  |Openembedded-core mailing list
>  |Openembedded-core@lists.openembedded.org
>  |http://lists.openembedded.org/mailman/listinfo/openembedded-core
>
>
>
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core
Ross Burton - Jan. 22, 2014, 6:06 p.m.
On 22 January 2014 17:56, Burton, Ross <ross.burton@intel.com> wrote:
> Whoops, this fell out of my inbox.   Whilst the is-active check now
> means that the service is restarted on upgrades, if the package is
> being installed it's now not started if AUTO_ENABLE is set to true.
>
> I think the logic should be "$D unset and (auto enable or is-active)"

Also, before doing systemctl reload, systemd needs to be told to
re-load the services with systemctl --system daemon-reload (thanks
Koen for pointing this out).

Ross

Patch

diff --git a/meta/classes/systemd.bbclass b/meta/classes/systemd.bbclass
index 3700b2e..1feac06 100644
--- a/meta/classes/systemd.bbclass
+++ b/meta/classes/systemd.bbclass
@@ -32,7 +32,8 @@  fi
 if type systemctl >/dev/null 2>/dev/null; then
 	systemctl $OPTS ${SYSTEMD_AUTO_ENABLE} ${SYSTEMD_SERVICE}
 -	if [ -z "$D" -a "${SYSTEMD_AUTO_ENABLE}" = "enable" ]; then
+	if [ -z "$D" ] && systemctl is-active ${SYSTEMD_SERVICE} &>/dev/null; then
+		# if service is active, restart it
 		systemctl restart ${SYSTEMD_SERVICE}
 	fi
 fi
--