On Fri, Apr 22, 2022 at 02:58:19PM +0200, Michal Hocko wrote: > On Thu 21-04-22 19:48:37, Kent Overstreet wrote: > > This patch: > > - Changes show_mem() to always report on slab usage > > - Instead of reporting on all slabs, we only report on top 10 slabs, > > and in sorted order > > As I've already pointed out in the email thread for the previous > version, this would be better in its own patch explaining why we want to > make this unconditional and why to limit the number caches to print. > Why the trashold shouldn't be absolute size based? > > > - Also reports on shrinkers, with the new shrinkers_to_text(). > > Shrinkers need to be included in OOM/allocation failure reporting > > because they're responsible for memory reclaim - if a shrinker isn't > > giving up its memory, we need to know which one and why. > > Again, I do agree that information about shrinkers can be useful but > there are two main things to consider. Do we want to dump that > information unconditionaly? E.g. does it make sense to print for all > allocation requests (even high order, GFP_NOWAIT...)? Should there be > any explicit trigger when to dump this data (like too many shrinkers > failing etc)? To add a concern: largest shrinkers are usually memcg-aware. Scanning over the whole cgroup tree (with potentially hundreds or thousands of cgroups) and over all shrinkers from the oom context sounds like a bad idea to me. IMO it's more appropriate to do from userspace by oomd or a similar daemon, well before the in-kernel OOM kicks in. > > Last but not least let me echo the concern from the other reply. Memory > allocations are not really reasonable to be done from the oom context so > the pr_buf doesn't sound like a good tool here. +1
