** Changed in: network-manager (Ubuntu Bionic) Importance: Undecided => Medium
-- You received this bug notification because you are a member of Ubuntu Touch seeded packages, which is subscribed to network-manager in Ubuntu. https://bugs.launchpad.net/bugs/1733321 Title: network-manager ADT tests fail with on ppc64el with artful/linux 4.13.0.17.18 Status in network-manager package in Ubuntu: New Status in network-manager source package in Artful: Won't Fix Status in network-manager source package in Bionic: Fix Committed Status in network-manager source package in Disco: Won't Fix Status in network-manager source package in Eoan: Fix Committed Status in network-manager source package in Focal: Fix Committed Bug description: [Impact] The killswitches-no-urfkill autopkgtest fails sometimes because nmcli reports the old state when it's called right after rfkill block/unblock. ppc64el ADT log from failed testcase: https://objectstorage.prodstack4-5.canonical.com/v1/AUTH_77e2ada1e7a84929a74ba3b87153c0ac /autopkgtest-artful/artful/ppc64el/n/network- manager/20171120_100719_28642@/log.gz Testcase output: ----- autopkgtest [10:04:48]: test killswitches-no-urfkill: [----------------------- make -C /lib/modules/4.13.0-17-generic/build KBUILD_SRC=/lib/modules/4.13.0-17-generic/build M=/tmp/autopkgtest.yE1hsA/build.62e/src/debian/tests make[1]: Entering directory '/usr/src/linux-headers-4.13.0-17-generic' AR /tmp/autopkgtest.yE1hsA/build.62e/src/debian/tests/built-in.o CC [M] /tmp/autopkgtest.yE1hsA/build.62e/src/debian/tests/fake-rfkill.o Building modules, stage 2. MODPOST 1 modules CC /tmp/autopkgtest.yE1hsA/build.62e/src/debian/tests/fake-rfkill.mod.o LD [M] /tmp/autopkgtest.yE1hsA/build.62e/src/debian/tests/fake-rfkill.ko make[1]: Leaving directory '/usr/src/linux-headers-4.13.0-17-generic' ERROR: NM could not track device state. autopkgtest [10:05:20]: test killswitches-no-urfkill: -----------------------] autopkgtest [10:05:20]: test killswitches-no-urfkill: - - - - - - - - - - results - - - - - - - - - - killswitches-no-urfkill FAIL non-zero exit status 1 ----- Package versions [artful/ppc64el]: network-manager 1.8.4-1ubuntu3 linux-meta 4.13.0.17.18 [Test Case] Assume that test vm and host are ppc64 1. deploy ppc64 vm instance ( with 1 cpu ) 2. modprobe mac80211_hwsim ( may need to install linux-modules-extra- pkg ) 3. apt install network-manager rfkill 4. modify /etc/netplan/[conf], renderer as NetworkManager 5. netplan apply 6. run below command - nmcli radio wifi ; rfkill list 0 ; rfkill block 0 ; rfkill list 0 ; nmcli radio wifi ; rfkill list 0 ; rfkill unblock 0 ; rfkill list 0 ; nmcli radio wifi enabled 0: fake: Wireless LAN Soft blocked: no Hard blocked: no 0: fake: Wireless LAN Soft blocked: yes Hard blocked: no enabled 0: fake: Wireless LAN Soft blocked: yes Hard blocked: no 0: fake: Wireless LAN Soft blocked: no Hard blocked: no enabled second 'enabled' should be 'disabled' but not updated properly. Adding "udevadm settle" in test file ( killswitches-no-urkfill ) between rfkill block/unblock and nmcli radio wifi will help updating status changes after rfkill block/unblock. [Regression Potential] This fixes testcase only, so any regression would cause incorrect test pass/fail, and might cause other missed bugs. [scope] this is needed for all releases. Debian does not include this autopkgtest, and so does not need this fix. [Other Info] this is caused by the 'systemd-rfkill.socket' listening to rfkill, and starting up 'systemd-rfkill.service' for any change. On a 1-cpu system (which all autopkgtest instances for network-manager are), this service startup sometimes blocks the uevent from reaching network- manager before the autopkgtest proceeds to call nmcli to check the rfkill status. This causes the test case to fail. There are (at least) 2 options to fix this: 1) stop/disable the 'systemd-rfkill.socket' at the start of the autopkgtest 2) call udevadm settle between the rfkill block/unblock and the nmcli call to check status This does not need to be fixed outside the test case, as normal nmcli use should be not done by users in a script immediately after changing rfkill state, nor is there anything that either rfkill or nmcli could even do to address this (technically, nmcli could internally issue a 'udevadm settle' every time it's called, but that seems paranoid and extreme). [original description] On ppc64el architecture, it was observed that a fix for systemd is also needed (see bug 1734908) for the testcase to be successful. # original test case 2.1. Download network-manager package source code $ apt-get source network-manager 2.2. Run killswitches-no-urfkill testcase $ cd network-manager-1.8.4 $ sudo ./debian/tests/killswitches-no-urfkill To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/network-manager/+bug/1733321/+subscriptions -- Mailing list: https://launchpad.net/~touch-packages Post to : touch-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~touch-packages More help : https://help.launchpad.net/ListHelp
