Bug 1685286

Description of problem:
Taking a closer look at the recent CVE-2019-6454 I was sending D-Bus messages with large path strings to PID 1.

While the patched package doesn't seem to crash anymore it seems to leak the memory associated with the messages.

Version-Release number of selected component (if applicable):

systemd-239-12.git8bca462.fc29

How reproducible:
Always

Steps to Reproduce:
1. Check PID1's memory usage:
 grep -i Rss /proc/1/status
 VmRSS:	    7960 kB
 RssAnon:	    4296 kB
 RssFile:	    3664 kB
 RssShmem:	       0 kB
2. Send invalid messages to PID1:
 for i in `seq 1 10`; do gdbus call -y -d org.freedesktop.systemd1 -m org.freedesktop.systemd1.Manager.StartUnit -o "/$(for x in `seq 0 28000`; do echo -n $x; done)"; done

Note that this takes some time as gdbus waits for a reply. This makes each message at least ~128KiB large due to the path string.

3.Check PID1's memory usage:
 grep -i Rss /proc/1/status
 VmRSS:	   10600 kB
 RssAnon:	    6936 kB
 RssFile:	    3664 kB
 RssShmem:	       0 kB


Actual results:
PID 1 drops the invalid D-Bus message, but doesn't seem to release all associated ressources.

Expected results:
PID 1 drops the invalid D-Bus message and releases all associated ressources.

Additional info:
PID1 nicely logs that it drops the messages if log level is set to 'debug':
systemd[1]: Received invalid message from connection bus-api-system, dropping.