I verified by creating a prerm script for packagekit-tools with sleep 20 in it, and then looked at the locks while that was sleeping:
# lslocks COMMAND PID TYPE SIZE MODE M START END PATH cron 415 FLOCK 4B WRITE 0 0 0 /run/crond.pid packagekitd 3374 POSIX 0B WRITE 0 0 0 /var/lib/dpkg/lock-frontend packagekitd 3374 POSIX 0B WRITE 0 0 0 /var/cache/apt/archives/lock atd 406 POSIX 4B WRITE 0 0 0 /run/atd.pid dpkg 3500 POSIX 0B WRITE 0 0 0 /var/lib/dpkg/lock We can see the packagekitd holds the frontend lock, and dpkg holds the lock. -- You received this bug notification because you are a member of Ubuntu Touch seeded packages, which is subscribed to packagekit in Ubuntu. https://bugs.launchpad.net/bugs/1795614 Title: packagekit frontend locking Status in packagekit package in Ubuntu: Fix Released Status in packagekit source package in Xenial: Fix Committed Status in packagekit source package in Bionic: Fix Committed Status in packagekit source package in Cosmic: Fix Released Bug description: [SRU Notes] [rbasak] Requires apt 1.2.28 (Xenial) and 1.6.5 (Bionic), both of which are currently in proposed. Do not release to updates without releasing the newer apt first. [Impact] PackageKit needs an adjustment for frontend locking, so it does not release the frontend lock during dpkg invocations, but only the normal dpkg lock. Frontend locking prevents race conditions between multiple dpkg frontends, which can cause other frontends to be interrupted while they are installing stuff - the frontend loses the dpkg lock between dpkg runs and the system ends up in a partially configured state. See bug 1781169 for more details on frontend locking. [Test case] 1. Install a package 2. Modify prerm to sleep 3. Remove package via pkcon and check that packagekitd process holds lock-frontend and dpkg holds lock (we release lock by closing the lock files, so just check if the files are open). [Regression potential] Changing the code to call UnLockInner() vs UnLock() makes it do less steps and only release "lock" as before, and not "lock-frontend". That should not be causing any regressions. The patch also adds a call to LockInner() after the dpkg execution to make it reacquire "lock", this could fail. It should not have much impact however, as it only affects a single transaction AFAICT. It does reduce the risk of some frontend not implementing frontend locking from racing while we still hold the frontend lock, though. [Other info] This is part of a wider series of SRUs for frontend locking - dpkg (bug 1796081) - apt (bug 1781169) - python-apt (bug 1795407) - packagekit (bug 1795614) - unattended-upgrades (bug 1789637) - aptdaemon (no bug filed yet) Further details about frontend locking can be found in https://lists.debian.org/debian-dpkg/2017/01/msg00044.html To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/packagekit/+bug/1795614/+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
