Hi Sergey,
On Mon, Apr 17, 2023 at 4:03 AM Sergey Senozhatsky
<senozhatsky@xxxxxxxxxxxx> wrote:
>
> zsmalloc pool can be compacted concurrently by many contexts,
> e.g.
>
> cc1 handle_mm_fault()
> do_anonymous_page()
> __alloc_pages_slowpath()
> try_to_free_pages()
> do_try_to_free_pages(
> lru_gen_shrink_node()
> shrink_slab()
> do_shrink_slab()
> zs_shrinker_scan()
> zs_compact()
>
> This creates unnecessary contention as all those processes
> compete for access to the same classes. A single compaction
> process is enough. Moreover contention that is created by
> multiple compaction processes impact other zsmalloc functions,
> e.g. zs_malloc(), since zsmalloc uses "global" pool->lock to
> synchronize access to pool.
>
> Introduce pool compaction mutex and permit only one compaction
> context at a time.
I am not sure what's the best practice here, but if the only use of
the mutex is a trylock, do we need a mutex here? It seems like a
simple atomic would do the trick. Perhaps something like:
static atomic_t ongoing_compaction = ATOMIC_INIT(0);
...
if (atomic_xchg(&ongoing_compaction, 1))
return;
....
atomic_set(&ongoing_compaction, 0);
FWIW, I am suggesting this because it was recently brought up when one
of my patches was reviewed that locks should protect data not code
paths [1], so I am trying to apply what I learned there :)
[1]https://lore.kernel.org/lkml/20230323172732.GE739026@xxxxxxxxxxx/
>
> /proc/lock-stat after make -j$((`nproc`+1)) linux kernel for
> &pool->lock#3:
>
> Base Patched
> --------------------------------------
> con-bounces 9797655 8125860
> contentions 11131185 9242153
> waittime-min 0.09 0.10
> waittime-max 4171695.76 3926258.74
> waittime-total 506197629.16 417061026.20
> waittime-avg 45.48 45.13
> acq-bounces 13809103 11383480
> acquisitions 21145155 18049364
> holdtime-min 0.06 0.07
> holdtime-max 7379928.80 3926274.89
> holdtime-total 46273950.89 37279624.53
> holdtime-avg 2.19 2.07
>
> Signed-off-by: Sergey Senozhatsky <senozhatsky@xxxxxxxxxxxx>
> ---
> mm/zsmalloc.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
> index cc81dfba05a0..0e036ec56c3c 100644
> --- a/mm/zsmalloc.c
> +++ b/mm/zsmalloc.c
> @@ -264,6 +264,7 @@ struct zs_pool {
> struct work_struct free_work;
> #endif
> spinlock_t lock;
> + struct mutex compact_lock;
> };
>
> struct zspage {
> @@ -2274,6 +2275,9 @@ unsigned long zs_compact(struct zs_pool *pool)
> struct size_class *class;
> unsigned long pages_freed = 0;
>
> + if (!mutex_trylock(&pool->compact_lock))
> + return 0;
> +
> for (i = ZS_SIZE_CLASSES - 1; i >= 0; i--) {
> class = pool->size_class[i];
> if (class->index != i)
> @@ -2281,6 +2285,7 @@ unsigned long zs_compact(struct zs_pool *pool)
> pages_freed += __zs_compact(pool, class);
> }
> atomic_long_add(pages_freed, &pool->stats.pages_compacted);
> + mutex_unlock(&pool->compact_lock);
>
> return pages_freed;
> }
> @@ -2388,6 +2393,7 @@ struct zs_pool *zs_create_pool(const char *name)
>
> init_deferred_free(pool);
> spin_lock_init(&pool->lock);
> + mutex_init(&pool->compact_lock);
>
> pool->name = kstrdup(name, GFP_KERNEL);
> if (!pool->name)
> --
> 2.40.0.634.g4ca3ef3211-goog
>
