On 05/09, Eric W. Biederman wrote: > > Oleg Nesterov <o...@redhat.com> writes: > > >> The patch does solve the issue. There should be nothing a userspace > >> process can observe that should tell it where in the middle of exec > >> such a migration happend so placing the migration at what from the > >> kernel's perspective might be technically later should not be a problem. > >> > >> If it is a problem the issue is that there is a way to observe the > >> difference. > > > > So. The task migrates from some MEMCG right after bprm_mm_init(). > > > > copy_strings() triggers OOM in MEMCG. This is quite possible, it can use a > > lot > > of memory and that is why we have acct_arg_size() to make these allocations > > visible to oom killer. > > > > task_in_mem_cgroup(MEMCG) returns false and oom killer has to kill another > > innocent process in MEMCG. > > > > Does this look like a way to observe the difference? > > Sort of. > > I don't know how the memcg gets away without migrating charges > when it migrates a process. With charges not being migrated > I don't think this is observable.
Not sure I understand how this connects to accounting... But yes sure, with or without your change, mem_cgroup_move_task() obviously can't see the the nascent bprm->mm. I have no idea if this is important or not, and iiuc cgroup v2 doesn't even support ->move_charge_at_immigrate. As for accounting, I still think that it would be better to nullify ->memcg in mm_init_memcg(), simply because we can not initialize it properly, we can race with migration until exec_mmap/cgroup_post_fork which need to update ->memcg anyway. Yes, this means a special case in get_mem_cgroup_from_mm(). Oleg. > That does look like a real issue however. > > >> > Perhaps we can change get_mem_cgroup_from_mm() to use > >> > mem_cgroup_from_css(current, memory_cgrp_id) if mm->memcg == NULL? > >> > >> Please God no. Having any unnecessary special case is just going to > >> confuse people and cause bugs. > > > > To me the unnecessary special case is the new_mm->memcg which is used for > > accounting but doesn't follow migration till exec_mmap(). But I won't > > argue. > > Eric
