On Thu, Nov 30, 2023 at 01:33:40PM -0800, Andrew Morton wrote: > On Thu, 30 Nov 2023 07:36:54 -0800 Dan Schatzberg <schatzberg.dan@xxxxxxxxx> wrote: > > > Allow proactive reclaimers to submit an additional swappiness=<val> > > argument to memory.reclaim. This overrides the global or per-memcg > > swappiness setting for that reclaim attempt. > > > > For example: > > > > echo "2M swappiness=0" > /sys/fs/cgroup/memory.reclaim > > > > will perform reclaim on the rootcg with a swappiness setting of 0 (no > > swap) regardless of the vm.swappiness sysctl setting. > > > > Signed-off-by: Dan Schatzberg <schatzberg.dan@xxxxxxxxx> > > --- > > include/linux/swap.h | 3 ++- > > mm/memcontrol.c | 55 +++++++++++++++++++++++++++++++++++--------- > > mm/vmscan.c | 13 +++++++++-- > > Documentation/admin-guide/cgroup-v2.rst is feeling unloved! Oops - total oversight on my part. I'll add docs in a V2 if we can come to consensus on this interface change in general. > > Please check whether this interface change can lead to > non-backward-compatible userspace. If someone's script does the above > echo command, will it break on older kernels? If so, does it matter? Older kernels will return -EINVAL with such a command - that seems appropriate, indicating that the argument is not supported.