[PATCH v2 bpf-next 08/12] bpf: Adjust low/high watermarks in bpf_mem_cache

[Alexei Starovoitov <alexei.starovoitov@gmail.com>]

From: Alexei Starovoitov <ast@kernel.org>

Same low/high watermarks for every bucket in bpf_mem_cache consume
significant amount of memory. Preallocating 64 elements of PAGE_SIZE
to the free list is not efficient.
Make low/high watermarks and batching value depend on element size.
This change brings significant memory savings.

Signed-off-by: Alexei Starovoitov <ast@kernel.org>
---
 kernel/bpf/memalloc.c | 64 ++++++++++++++++++++++++++++++-------------
 1 file changed, 45 insertions(+), 19 deletions(-)

diff --git a/kernel/bpf/memalloc.c b/kernel/bpf/memalloc.c
index a43630371b9f..be8262f5c9ec 100644
--- a/kernel/bpf/memalloc.c
+++ b/kernel/bpf/memalloc.c
@@ -105,6 +105,7 @@ struct bpf_mem_cache {
 	atomic_t free_cnt_nmi;
 	/* flag to refill nmi list too */
 	bool refill_nmi_list;
+	int low_watermark, high_watermark, batch;
 };
 
 struct bpf_mem_caches {
@@ -123,14 +124,6 @@ static struct llist_node notrace *__llist_del_first(struct llist_head *head)
 	return entry;
 }
 
-#define BATCH 48
-#define LOW_WATERMARK 32
-#define HIGH_WATERMARK 96
-/* Assuming the average number of elements per bucket is 64, when all buckets
- * are used the total memory will be: 64*16*32 + 64*32*32 + 64*64*32 + ... +
- * 64*4096*32 ~ 20Mbyte
- */
-
 /* extra macro useful for testing by randomizing in_nmi condition */
 #define bpf_in_nmi() in_nmi()
 
@@ -238,7 +231,7 @@ static void free_bulk(struct bpf_mem_cache *c)
 		if (IS_ENABLED(CONFIG_PREEMPT_RT))
 			local_irq_restore(flags);
 		free_one(c, llnode);
-	} while (cnt > (HIGH_WATERMARK + LOW_WATERMARK) / 2);
+	} while (cnt > (c->high_watermark + c->low_watermark) / 2);
 }
 
 static void free_bulk_nmi(struct bpf_mem_cache *c)
@@ -253,7 +246,7 @@ static void free_bulk_nmi(struct bpf_mem_cache *c)
 		else
 			cnt = 0;
 		free_one(c, llnode);
-	} while (cnt > (HIGH_WATERMARK + LOW_WATERMARK) / 2);
+	} while (cnt > (c->high_watermark + c->low_watermark) / 2);
 }
 
 static void bpf_mem_refill(struct irq_work *work)
@@ -262,12 +255,12 @@ static void bpf_mem_refill(struct irq_work *work)
 	int cnt;
 
 	cnt = c->free_cnt;
-	if (cnt < LOW_WATERMARK)
+	if (cnt < c->low_watermark)
 		/* irq_work runs on this cpu and kmalloc will allocate
 		 * from the current numa node which is what we want here.
 		 */
-		alloc_bulk(c, BATCH, NUMA_NO_NODE);
-	else if (cnt > HIGH_WATERMARK)
+		alloc_bulk(c, c->batch, NUMA_NO_NODE);
+	else if (cnt > c->high_watermark)
 		free_bulk(c);
 
 	if (!c->refill_nmi_list)
@@ -276,9 +269,9 @@ static void bpf_mem_refill(struct irq_work *work)
 		 */
 		return;
 	cnt = atomic_read(&c->free_cnt_nmi);
-	if (cnt < LOW_WATERMARK)
-		alloc_bulk_nmi(c, BATCH, NUMA_NO_NODE);
-	else if (cnt > HIGH_WATERMARK)
+	if (cnt < c->low_watermark)
+		alloc_bulk_nmi(c, c->batch, NUMA_NO_NODE);
+	else if (cnt > c->high_watermark)
 		free_bulk_nmi(c);
 	c->refill_nmi_list = false;
 }
@@ -294,14 +287,47 @@ static void notrace irq_work_raise(struct bpf_mem_cache *c, bool in_nmi)
 	irq_work_queue(&c->refill_work);
 }
 
+/* For typical bpf map case that uses bpf_mem_cache_alloc and single bucket
+ * the freelist cache will be elem_size * 64 (or less) on each cpu.
+ *
+ * For bpf programs that don't have statically known allocation sizes and
+ * assuming (low_mark + high_mark) / 2 as an average number of elements per
+ * bucket and all buckets are used the total amount of memory in freelists
+ * on each cpu will be:
+ * 64*16 + 64*32 + 64*64 + 64*96 + 64*128 + 64*196 + 64*256 + 32*512 + 16*1024 + 8*2048 + 4*4096
+ * + nmi's reserves
+ * 1*16 + 1*32 + 1*64 + 1*96 + 1*128 + 1*196 + 1*256 + 1*512 + 1*1024 + 1*2048 + 1*4096
+ * == ~ 122 Kbyte using below heuristic.
+ * In unlikely worst case where bpf progs used all allocations sizes from
+ * non-NMI and from NMI too: ~ 227 Kbyte per cpu.
+ * Initialized, but unused bpf allocator (not bpf map specific one) will
+ * consume ~ 19 Kbyte per cpu.
+ * Typical case will be between 19K and 122K closer to 19K.
+ * bpf progs can and should share bpf_mem_cache when possible.
+ */
+
 static void prefill_mem_cache(struct bpf_mem_cache *c, int cpu)
 {
 	init_irq_work(&c->refill_work, bpf_mem_refill);
+	if (c->unit_size <= 256) {
+		c->low_watermark = 32;
+		c->high_watermark = 96;
+	} else {
+		/* When page_size == 4k, order-0 cache will have low_mark == 2
+		 * and high_mark == 6 with batch alloc of 3 individual pages at
+		 * a time.
+		 * 8k allocs and above low == 1, high == 3, batch == 1.
+		 */
+		c->low_watermark = max(32 * 256 / c->unit_size, 1);
+		c->high_watermark = max(96 * 256 / c->unit_size, 3);
+	}
+	c->batch = max((c->high_watermark - c->low_watermark) / 4 * 3, 1);
+
 	/* To avoid consuming memory assume that 1st run of bpf
 	 * prog won't be doing more than 4 map_update_elem from
 	 * irq disabled region
 	 */
-	alloc_bulk(c, c->unit_size < 256 ? 4 : 1, cpu_to_node(cpu));
+	alloc_bulk(c, c->unit_size <= 256 ? 4 : 1, cpu_to_node(cpu));
 
 	/* NMI progs are rare. Assume they have one map_update
 	 * per prog at the very beginning.
@@ -442,7 +468,7 @@ static void notrace *unit_alloc(struct bpf_mem_cache *c)
 	}
 	WARN_ON(cnt < 0);
 
-	if (cnt < LOW_WATERMARK)
+	if (cnt < c->low_watermark)
 		irq_work_raise(c, in_nmi);
 	return llnode;
 }
@@ -471,7 +497,7 @@ static void notrace unit_free(struct bpf_mem_cache *c, void *ptr)
 	}
 	WARN_ON(cnt <= 0);
 
-	if (cnt > HIGH_WATERMARK)
+	if (cnt > c->high_watermark)
 		/* free few objects from current cpu into global kmalloc pool */
 		irq_work_raise(c, in_nmi);
 }
-- 
2.30.2