memcg: add flush tracepoint

This tracepoint gives visibility on how often the flushing of memcg stats
occurs and contains info on whether it was forced, skipped, and the value
of stats updated.  It can help with understanding how readers are affected
by having to perform the flush, and the effectiveness of the flush by
inspecting the number of stats updated.  Paired with the recently added
tracepoints for tracing rstat updates, it can also help show correlation
where stats exceed thresholds frequently.

Link: https://lkml.kernel.org/r/20241029021106.25587-3-inwardvessel@gmail.com
Signed-off-by: JP Kobryn <inwardvessel@gmail.com>
Reviewed-by: Yosry Ahmed <yosryahmed@google.com>
Acked-by: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

diff --git a/include/trace/events/memcg.h b/include/trace/events/memcg.h
index 8667e57816d213972165815cdc1546ee56f6d406..dfe2f51019b4c126064759d68c23108644b52ba8 100644 (file)
--- a/include/trace/events/memcg.h
+++ b/include/trace/events/memcg.h
@@ -74,6 +74,31 @@ DEFINE_EVENT(memcg_rstat_events, count_memcg_events,
        TP_ARGS(memcg, item, val)
 );
 
+TRACE_EVENT(memcg_flush_stats,
+
+       TP_PROTO(struct mem_cgroup *memcg, s64 stats_updates,
+               bool force, bool needs_flush),
+
+       TP_ARGS(memcg, stats_updates, force, needs_flush),
+
+       TP_STRUCT__entry(
+               __field(u64, id)
+               __field(s64, stats_updates)
+               __field(bool, force)
+               __field(bool, needs_flush)
+       ),
+
+       TP_fast_assign(
+               __entry->id = cgroup_id(memcg->css.cgroup);
+               __entry->stats_updates = stats_updates;
+               __entry->force = force;
+               __entry->needs_flush = needs_flush;
+       ),
+
+       TP_printk("memcg_id=%llu stats_updates=%lld force=%d needs_flush=%d",
+               __entry->id, __entry->stats_updates,
+               __entry->force, __entry->needs_flush)
+);
 
 #endif /* _TRACE_MEMCG_H */
 

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index a4da834b4aa32ea9b41aff2d1d397e0e3c784166..6486e96528434fc7134dee79d4cc3965451a266b 100644 (file)
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -597,7 +597,12 @@ static inline void memcg_rstat_updated(struct mem_cgroup *memcg, int val)
 
 static void __mem_cgroup_flush_stats(struct mem_cgroup *memcg, bool force)
 {
-       if (!force && !memcg_vmstats_needs_flush(memcg->vmstats))
+       bool needs_flush = memcg_vmstats_needs_flush(memcg->vmstats);
+
+       trace_memcg_flush_stats(memcg, atomic64_read(&memcg->vmstats->stats_updates),
+               force, needs_flush);
+
+       if (!force && !needs_flush)
                return;
 
        if (mem_cgroup_is_root(memcg))