On Sun, 23 Feb 2020, Hugh Dickins wrote: > On Thu, 20 Feb 2020, Dan Schatzberg wrote: > > > memalloc_use_memcg() worked for kernel allocations but was silently > > ignored for user pages. > > > > This patch establishes a precedence order for who gets charged: > > > > 1. If there is a memcg associated with the page already, that memcg is > > charged. This happens during swapin. > > > > 2. If an explicit mm is passed, mm->memcg is charged. This happens > > during page faults, which can be triggered in remote VMs (eg gup). > > > > 3. Otherwise consult the current process context. If it has configured > > a current->active_memcg, use that. Otherwise, current->mm->memcg. > > > > Previously, if a NULL mm was passed to mem_cgroup_try_charge (case 3) it > > would always charge the root cgroup. Now it looks up the current > > active_memcg first (falling back to charging the root cgroup if not > > set). > > > > Signed-off-by: Dan Schatzberg <schatzberg.dan@xxxxxxxxx> > > Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx> > > Acked-by: Tejun Heo <tj@xxxxxxxxxx> > > Acked-by: Hugh Dickins <hughd@xxxxxxxxxx> > > Yes, internally we have some further not-yet-upstreamed complications > here (mainly, the "memcg=" mount option for all charges on a tmpfs to > be charged to that memcg); but what you're doing here does not obstruct > adding that later, they fit in well with the hierarchy that you (and > Johannes) mapped out above, and it's really an improvement for shmem > not to be referring to current there - thanks. I acked slightly too soon. There are two other uses of "try_charge" in mm/shmem.c: we can be confident that the userfaultfd one knows what mm it's dealing with, but the shmem_swapin_page() instance has a similar use of current->mm, that you also want to adjust to NULL, don't you? Hugh
