On Wed, Jun 26, 2024 at 2:18 PM Jesper Dangaard Brouer <hawk@xxxxxxxxxx> wrote:
>
> Avoid lock contention on the global cgroup rstat lock caused by kswapd
> starting on all NUMA nodes simultaneously. At Cloudflare, we observed
> massive issues due to kswapd and the specific mem_cgroup_flush_stats()
> call inlined in shrink_node, which takes the rstat lock.
>
> On our 12 NUMA node machines, each with a kswapd kthread per NUMA node,
> we noted severe lock contention on the rstat lock. This contention
> causes 12 CPUs to waste cycles spinning every time kswapd runs.
> Fleet-wide stats (/proc/N/schedstat) for kthreads revealed that we are
> burning an average of 20,000 CPU cores fleet-wide on kswapd, primarily
> due to spinning on the rstat lock.
>
> To help reviewer follow code: When the Per-CPU-Pages (PCP) freelist is
> empty, __alloc_pages_slowpath calls wake_all_kswapds(), causing all
> kswapdN threads to wake up simultaneously. The kswapd thread invokes
> shrink_node (via balance_pgdat) triggering the cgroup rstat flush
> operation as part of its work. This results in kernel self-induced rstat
> lock contention by waking up all kswapd threads simultaneously.
> Leveraging this detail: balance_pgdat() have NULL value in
> target_mem_cgroup, this cause mem_cgroup_flush_stats() to do flush with
> root_mem_cgroup.
>
> To avoid this kind of thundering herd problem, kernel previously had a
> "stats_flush_ongoing" concept, but this was removed as part of commit
> 7d7ef0a4686a ("mm: memcg: restore subtree stats flushing"). This patch
> reintroduce and generalized the concept to apply to all users of cgroup
> rstat, not just memcg.
>
> If there is an ongoing rstat flush, and current cgroup is a descendant,
> then it is unnecessary to do the flush. For callers to still see updated
> stats, wait for ongoing flusher to complete before returning, but add
> timeout as stats are already inaccurate given updaters keeps running.
>
> Fixes: 7d7ef0a4686a ("mm: memcg: restore subtree stats flushing").
> Signed-off-by: Jesper Dangaard Brouer <hawk@xxxxxxxxxx>
> ---
> V2: https://lore.kernel.org/all/171923011608.1500238.3591002573732683639.stgit@firesoul/
> V1: https://lore.kernel.org/all/171898037079.1222367.13467317484793748519.stgit@firesoul/
> RFC: https://lore.kernel.org/all/171895533185.1084853.3033751561302228252.stgit@firesoul/
>
> kernel/cgroup/rstat.c | 61 ++++++++++++++++++++++++++++++++++++++++---------
> 1 file changed, 50 insertions(+), 11 deletions(-)
>
> diff --git a/kernel/cgroup/rstat.c b/kernel/cgroup/rstat.c
> index 2a42be3a9bb3..f21e6b1109a4 100644
> --- a/kernel/cgroup/rstat.c
> +++ b/kernel/cgroup/rstat.c
> @@ -2,6 +2,7 @@
> #include "cgroup-internal.h"
>
> #include <linux/sched/cputime.h>
> +#include <linux/completion.h>
>
> #include <linux/bpf.h>
> #include <linux/btf.h>
> @@ -11,6 +12,8 @@
>
> static DEFINE_SPINLOCK(cgroup_rstat_lock);
> static DEFINE_PER_CPU(raw_spinlock_t, cgroup_rstat_cpu_lock);
> +static struct cgroup *cgrp_rstat_ongoing_flusher;
> +static DECLARE_COMPLETION(cgrp_rstat_flusher_done);
>
> static void cgroup_base_stat_flush(struct cgroup *cgrp, int cpu);
>
> @@ -346,6 +349,44 @@ static void cgroup_rstat_flush_locked(struct cgroup *cgrp)
> }
> }
>
> +#define MAX_WAIT msecs_to_jiffies(100)
> +/* Trylock helper that also checks for on ongoing flusher */
> +static bool cgroup_rstat_trylock_flusher(struct cgroup *cgrp)
> +{
> +retry:
> + bool locked = __cgroup_rstat_trylock(cgrp, -1);
> + if (!locked) {
> + struct cgroup *cgrp_ongoing;
> +
> + /* Lock is contended, lets check if ongoing flusher is already
> + * taking care of this, if we are a descendant.
> + */
> + cgrp_ongoing = READ_ONCE(cgrp_rstat_ongoing_flusher);
> + if (!cgrp_ongoing)
> + goto retry;
> +
> + if (cgroup_is_descendant(cgrp, cgrp_ongoing)) {
> + wait_for_completion_interruptible_timeout(
> + &cgrp_rstat_flusher_done, MAX_WAIT);
Thanks for sending this!
The reason why I suggested that the completion live in struct cgroup
is because there is a chance here that the flush completes and another
irrelevant flush starts between reading cgrp_rstat_ongoing_flusher and
calling wait_for_completion_interruptible_timeout().
This will cause the caller to wait for an irrelevant flush, which may
be fine because today the caller would wait for the lock anyway. Just
mentioning this in case you think this may happen enough to be a
problem.
Also, I like the idea of the timeout here, it bounds the flush wait
time. I am wondering if there's a way to log something when the
timeout is exceeded (which probably means flushing is taking too
long), or maybe have a debug counter if we suspect this may spam the
log.
