Frank loaned me a test system. The test system in the "idle" state is enough to see the issue, with a persistent load of 3+ when nothing is happening except event loop spin. The test system is using a dasd as the disk that it is running from.
With some udev tracing, one can see an apparent loop, with udisks2 opening /dev/dasda, which triggers an inotify, which triggers udev rules, which trigger udisks2 again, or something similar to this flow.
Spent some time analyzing this.
Frank loaned me a test system. The test system in the "idle" state is enough to see the issue, with a persistent load of 3+ when nothing is happening except event loop spin. The test system is using a dasd as the disk that it is running from.
With some udev tracing, one can see an apparent loop, with udisks2 opening /dev/dasda, which triggers an inotify, which triggers udev rules, which trigger udisks2 again, or something similar to this flow.
Workarounds include: ="add|change" , KERNEL=="dasd*", OPTIONS:="nowatch"'
* adding a nowatch rule 'ACTION=
* systemctl stop udisks2
I've begun exploring udev changes for udisks2 to ignore dasd.