Re: [PATCH] mm, memcg: fix wrong mem cgroup protection

On Fri, Apr 24, 2020 at 9:44 PM Johannes Weiner hannes@cmpxchg.org wrote:

On Fri, Apr 24, 2020 at 09:14:52AM -0400, Johannes Weiner wrote:

...

However, mem_cgroup_protected() never expected anybody to look at the effective protection values when it indicated that the cgroup is above its protection. As a result, a query during limit reclaim may return stale protection values that were calculated by a previous reclaim cycle in which the cgroup did have siblings.

Btw, I think there is opportunity to make this a bit less error prone.

We have a mem_cgroup_protected() that returns yes or no, essentially, but protection isn't a binary state anymore.

It's also been a bit iffy that it looks like a simple predicate function, but it indeed needs to run procedurally for each cgroup in order for the calculations throughout the tree to be correct.

It might be better to have a

    mem_cgroup_calculate_protection()

that runs for every cgroup we visit and sets up the internal state; then have more self-explanatory query functions on top of that:

    mem_cgroup_below_min()
    mem_cgroup_below_low()
    mem_cgroup_protection()

What do you guys think?

diff --git a/mm/vmscan.c b/mm/vmscan.c index e0f502b5fca6..dbd3f75d39b9 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -2615,14 +2615,15 @@ static void shrink_node_memcgs(pg_data_t *pgdat, struct scan_control *sc) unsigned long reclaimed; unsigned long scanned;

•           switch (mem_cgroup_protected(target_memcg, memcg)) {
  

•           case MEMCG_PROT_MIN:
  

•           mem_cgroup_calculate_protection(target_memcg, memcg);
  

•           if (mem_cgroup_below_min(memcg)) {
                    /*
                     * Hard protection.
                     * If there is no reclaimable memory, OOM.
                     */
                    continue;
  

•           case MEMCG_PROT_LOW:
  

•           } else if (mem_cgroup_below_low(memcg)) {
                    /*
                     * Soft protection.
                     * Respect the protection only as long as
  

@@ -2634,16 +2635,6 @@ static void shrink_node_memcgs(pg_data_t *pgdat, struct scan_control *sc) continue; } memcg_memory_event(memcg, MEMCG_LOW);

•                   break;
  

•           case MEMCG_PROT_NONE:
  

•                   /*
  

•                    * All protection thresholds breached. We may
  

•                    * still choose to vary the scan pressure
  

•                    * applied based on by how much the cgroup in
  

•                    * question has exceeded its protection
  

•                    * thresholds (see get_scan_count).
  

•                    */
  

•                   break;
            }
  
            reclaimed = sc->nr_reclaimed;
  

After my revist of the memcg protection. I have another idea.

The emin and elow is not decided by the memcg(struct mem_cgroup), but they are really decided by the reclaim context(struct srhink_control). So they should not be bound into struct mem_cgroup, while they are really should be bound into struct srhink_control. IOW, we should move emin and elow from struct mem_cgroup into struct srhink_control. And they two members in shrink_control will be updated when a new memcg is to be shrinked. I haven't thought it deeply, but I think this should be the right thing to do.

Thanks Yafang