Hi All

I'm afraid epitaph has a point here. I've been doing some more digging into how the N900 manages resources and it seems that Open Hardware Manager is in use on our system.

OHM includes cgroup management, and it creates a cgroup filesystem in "/syspart".

Look at the contents of this folder on an un-patched device - they're almost identical to the /dev/cgroup/cpu folder contents on a patched device.

Have a look in /syspart on a patched device - there is no cgroup filesystem there anymore, it's just been moved to /dev/cgroup/cpu by the patch script.

OHM keeps its config files in /etc/ohm and /usr/share/policy/etc/current/. A look around the files in these directories will show you listings of every maemo process and what cgroup they belong in (applications, desktop, launcher...). This is the source of the mysteriously self-organising cgroup directory.

It's not mysterious, its not self-organising, OHM is doing its job as it's been configured to.

The patch script does make one change to the way cgroups are managed by OHM though. OHM manages CPU, memory and "freezer" state for all processes. The patch script only manages CPU and memory.

On an unpatched device, cat /syspart/<cgroup_name>/freezer.state will show you the current state of a cgroups processes.
I've only ever seen cgroups in the "THAWED" state on an N900, so I doubt they ever get frozen unless the device is in standby.

On an unptched device, there are no freezer.state files in /dev/cgroup. Not surprising as the mount of /dev/cgroup/cpu uses options "-o cpu, memory" but not "freezer".