On Mon 21-09-20 16:41:34, Christian Brauner wrote: > On Mon, Sep 21, 2020 at 03:42:00PM +0200, Michal Hocko wrote: > > [Cc Tejun and Christian - this is a part of a larger discussion which is > > not directly related to this particular question so let me trim the > > original email to the bare minimum.] > > > > On Fri 18-09-20 12:40:32, Peter Xu wrote: > > [...] > > > One issue is when we charge for cgroup we probably can't do that onto the new > > > mm/task, since copy_namespaces() is called after copy_mm(). I don't know > > > enough about cgroup, I thought the child will inherit the parent's, but I'm not > > > sure. Or, can we change that order of copy_namespaces() && copy_mm()? I don't > > > see a problem so far but I'd like to ask first.. > > > > I suspect you are referring to CLONE_INTO_CGROUP, right? I have only now > > learned about this feature so I am not deeply familiar with all the > > details and I might be easily wrong. Normally all the cgroup aware > > resources are accounted to the parent's cgroup. For memcg that includes > > all the page tables, early CoW and other allocations with __GFP_ACCOUNT. > > IIUC CLONE_INTO_CGROUP properly then this hasn't changed as the child is > > associated to its new cgroup (and memcg) only in cgroup_post_fork. If > > that is correct then we might have quite a lot of resources bound to > > child's lifetime but accounted to the parent's memcg which can lead to > > all sorts of interesting problems (e.g. unreclaimable memory - even by > > the oom killer). > > > > Christian, Tejun is this the expected semantic or I am just misreading > > the code? > > Hey Michal, > > Thanks for the Cc! > > If I understand your question correctly, then you are correct. The logic > is split in three simple parts: > 1. Child gets created and doesn't live in any cset > - This should mean that resources are still charged against the > parent's memcg which is what you're asking afiu. > 1. cgroup_can_fork() > - create new or find existing matching cset for the child > 3. cgroup_post_fork() > - move/attach child to the new or found cset > > _Purely from a CLONE_INTO_CGROUP perspective_ you should be ok to > reverse the order of copy_mm() and copy_namespaces(). Switching the order wouldn't make much of a difference right. At least not for memcg where all the accounted allocations will still go to current's memcg. -- Michal Hocko SUSE Labs
