On Mon, 2011-09-19 at 13:26 -0300, Glauber Costa wrote:Fair request.On 09/19/2011 01:19 PM, Peter Zijlstra wrote:On Wed, 2011-09-14 at 17:04 -0300, Glauber Costa wrote:Because this is the whole purpose of this series. Think that in aBesides user and system.
Why!?
reasonable use case, each cgroup will be tied to a container system.
Each of them wants to have its own statistics, in /proc/stat. The best
way I thought of doing it was to piggyback into cpuacct, instead of
coming up with a 124th accounting mechanism just for that.
I also used it do demonstrate how can we easily implement a new feature
found in hypervisors out of it, but this is really just the cherry on top.
OK, but then amend the changelog with something like:
Since we want to provide a full /proc/stat equivalent for cgroups, also
implement nice time.