On 09/07/2016 08:14 PM, Pau Espin Pedrol wrote:

2016-09-07 11:22 GMT+02:00 Chen Qi <[email protected] <mailto:[email protected]>>:

    Add new variable SYSTEMD_USER_SERVICE and SYSTEM_USER_AUTO_ENABLE
    to manage user services. Their usage is like SYSTEMD_SERVICE and
    SYSTEMD_AUTO_ENABLE.

    [YOCTO #7800]

    Signed-off-by: Chen Qi <[email protected]
    <mailto:[email protected]>>
    ---
     meta/classes/systemd.bbclass | 17 +++++++++++++----
     1 file changed, 13 insertions(+), 4 deletions(-)

    diff --git a/meta/classes/systemd.bbclass
    b/meta/classes/systemd.bbclass
    index db7873f..78cec97 100644
    --- a/meta/classes/systemd.bbclass
    +++ b/meta/classes/systemd.bbclass
    @@ -7,6 +7,7 @@ SYSTEMD_PACKAGES_class-nativesdk ?= ""

     # Whether to enable or disable the services on installation.
     SYSTEMD_AUTO_ENABLE ??= "enable"
    +SYSTEMD_USER_AUTO_ENABLE ??= "enable"

     # This class will be included in any recipe that supports systemd
    init scripts,
     # even if systemd is not in DISTRO_FEATURES.  As such don't make
    any changes
    @@ -30,10 +31,14 @@ fi

     if type systemctl >/dev/null 2>/dev/null; then
            systemctl $OPTS ${SYSTEMD_AUTO_ENABLE} ${SYSTEMD_SERVICE}
    +       systemctl $OPTS --global ${SYSTEMD_USER_AUTO_ENABLE}
    ${SYSTEMD_USER_SERVICE}


I'm not sure having these 2 systemctl being executed together everytime is a good idea. What if a recipe has a user service and no system service?

Hi Pau Espin Pedrol,

Thanks for your review.

The postinstall script runs successfully with expected result at rootfs time. But your question reminds me of the situation of the on-target install/remove situation. I think I'll need to make a new patch to make sure things work in both situations.

The key point here is that 'systemctl' at rootfs time is a shell script written by ourselves and 'systemctl' on target is that provided by systemd.

We are calling the first one with an empty system service? Or we are may be enabling a systemd system service which should not be enabled according to the recipe? I have the feeling this kind of cases are not being catch in here and other pkg scripts in this commit.

It is far from perfect, but in case you didn't, you may want to have a look at my initial/previous commit to try to fix the issue: https://www.mail-archive.com/[email protected]/msg42187.html


            if [ -z "$D" -a "${SYSTEMD_AUTO_ENABLE}" = "enable" ]; then
                    systemctl restart ${SYSTEMD_SERVICE}
            fi
    +       if [ -z "$D" -a "${SYSTEMD_USER_AUTO_ENABLE}" = "enable"
    ]; then
    +               systemctl --global restart ${SYSTEMD_USER_SERVICE}
    +       fi
     fi
     }

    @@ -47,9 +52,11 @@ fi
     if type systemctl >/dev/null 2>/dev/null; then
            if [ -z "$D" ]; then
                    systemctl stop ${SYSTEMD_SERVICE}
    +               systemctl --global stop ${SYSTEMD_USER_SERVICE}


I think this is not gonna work, you cannot call --global with stop afair. I'm not sure which is the good solution for this though, but you should ideally go through all systemd user sessions and call "systemctl --user stop". No idea if that's actually easily feasible.

The service is enabled and started with '--global' option, and stopping it with '--global' option seems reasonable to me. If there's some special case, let's look into it then.

Best Regards,
Chen Qi

            fi

            systemctl $OPTS disable ${SYSTEMD_SERVICE}
    +       systemctl $OPTS --global disable ${SYSTEMD_USER_SERVICE}
     fi
     }

    @@ -139,12 +146,14 @@ python systemd_populate_packages() {
         def systemd_check_services():
             searchpaths = [oe.path.join(d.getVar("sysconfdir", True),
    "systemd", "system"),]
             searchpaths.append(d.getVar("systemd_system_unitdir", True))
    +        searchpaths.append(oe.path.join(d.getVar("sysconfdir",
    True), "systemd", "user"))
    +        searchpaths.append(d.getVar("systemd_user_unitdir", True))
             systemd_packages = d.getVar('SYSTEMD_PACKAGES', True)

             keys = 'Also'
             # scan for all in SYSTEMD_SERVICE[]
             for pkg_systemd in systemd_packages.split():
    -            for service in get_package_var(d, 'SYSTEMD_SERVICE',
    pkg_systemd).split():
    +            for service in (get_package_var(d, 'SYSTEMD_SERVICE',
    pkg_systemd) + get_package_var(d, 'SYSTEMD_USER_SERVICE',
    pkg_systemd)).split():
                     path_found = ''

                     # Deal with adding, for example,
    '[email protected]' from
    @@ -165,14 +174,14 @@ python systemd_populate_packages() {
                     if path_found != '':
                         systemd_add_files_and_parse(pkg_systemd,
    path_found, service, keys)
                     else:
    -                    raise bb.build.FuncFailed("SYSTEMD_SERVICE_%s
    value %s does not exist" % \
    -                        (pkg_systemd, service))
    +                    raise bb.build.FuncFailed("SYSTEMD_SERVICE_%s
    or SYSTEMD_USER_SERVICE_%s value %s does not exist" % \
    +                        (pkg_systemd, pkg_systemd, service))

         # Run all modifications once when creating package
         if os.path.exists(d.getVar("D", True)):
             for pkg in d.getVar('SYSTEMD_PACKAGES', True).split():
                 systemd_check_package(pkg)
    -            if d.getVar('SYSTEMD_SERVICE_' + pkg, True):
    +            if d.getVar('SYSTEMD_SERVICE_' + pkg, True) or
    d.getVar('SYSTEMD_USER_SERVICE_' + pkg, True):
                     systemd_generate_package_scripts(pkg)
             systemd_check_services()
     }
    --
    1.9.1

    --
    _______________________________________________
    Openembedded-core mailing list
    [email protected]
    <mailto:[email protected]>
    http://lists.openembedded.org/mailman/listinfo/openembedded-core
    <http://lists.openembedded.org/mailman/listinfo/openembedded-core>



-- 
_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.openembedded.org/mailman/listinfo/openembedded-core

Reply via email to