1 files changed, 112 insertions, 61 deletions

diff --git a/include/net/fq_impl.h b/include/net/fq_impl.h
index 38a9a3d1222b..524b510f1c68 100644
--- a/include/net/fq_impl.h
+++ b/include/net/fq_impl.h
@@ -11,35 +11,37 @@
 
 /* functions that are embedded into includer */
 
-static void fq_adjust_removal(struct fq *fq,
-			      struct fq_flow *flow,
-			      struct sk_buff *skb)
+
+static void
+__fq_adjust_removal(struct fq *fq, struct fq_flow *flow, unsigned int packets,
+		    unsigned int bytes, unsigned int truesize)
 {
 	struct fq_tin *tin = flow->tin;
+	int idx;
 
-	tin->backlog_bytes -= skb->len;
-	tin->backlog_packets--;
-	flow->backlog -= skb->len;
-	fq->backlog--;
-	fq->memory_usage -= skb->truesize;
-}
+	tin->backlog_bytes -= bytes;
+	tin->backlog_packets -= packets;
+	flow->backlog -= bytes;
+	fq->backlog -= packets;
+	fq->memory_usage -= truesize;
 
-static void fq_rejigger_backlog(struct fq *fq, struct fq_flow *flow)
-{
-	struct fq_flow *i;
+	if (flow->backlog)
+		return;
 
-	if (flow->backlog == 0) {
-		list_del_init(&flow->backlogchain);
-	} else {
-		i = flow;
+	if (flow == &tin->default_flow) {
+		list_del_init(&tin->tin_list);
+		return;
+	}
 
-		list_for_each_entry_continue(i, &fq->backlogs, backlogchain)
-			if (i->backlog < flow->backlog)
-				break;
+	idx = flow - fq->flows;
+	__clear_bit(idx, fq->flows_bitmap);
+}
 
-		list_move_tail(&flow->backlogchain,
-			       &i->backlogchain);
-	}
+static void fq_adjust_removal(struct fq *fq,
+			      struct fq_flow *flow,
+			      struct sk_buff *skb)
+{
+	__fq_adjust_removal(fq, flow, 1, skb->len, skb->truesize);
 }
 
 static struct sk_buff *fq_flow_dequeue(struct fq *fq,
@@ -54,11 +56,37 @@ static struct sk_buff *fq_flow_dequeue(struct fq *fq,
 		return NULL;
 
 	fq_adjust_removal(fq, flow, skb);
-	fq_rejigger_backlog(fq, flow);
 
 	return skb;
 }
 
+static int fq_flow_drop(struct fq *fq, struct fq_flow *flow,
+			fq_skb_free_t free_func)
+{
+	unsigned int packets = 0, bytes = 0, truesize = 0;
+	struct fq_tin *tin = flow->tin;
+	struct sk_buff *skb;
+	int pending;
+
+	lockdep_assert_held(&fq->lock);
+
+	pending = min_t(int, 32, skb_queue_len(&flow->queue) / 2);
+	do {
+		skb = __skb_dequeue(&flow->queue);
+		if (!skb)
+			break;
+
+		packets++;
+		bytes += skb->len;
+		truesize += skb->truesize;
+		free_func(fq, tin, flow, skb);
+	} while (packets < pending);
+
+	__fq_adjust_removal(fq, flow, packets, bytes, truesize);
+
+	return packets;
+}
+
 static struct sk_buff *fq_tin_dequeue(struct fq *fq,
 				      struct fq_tin *tin,
 				      fq_tin_dequeue_t dequeue_func)
@@ -108,15 +136,14 @@ begin:
 
 static u32 fq_flow_idx(struct fq *fq, struct sk_buff *skb)
 {
-	u32 hash = skb_get_hash_perturb(skb, &fq->perturbation);
+	u32 hash = skb_get_hash(skb);
 
 	return reciprocal_scale(hash, fq->flows_cnt);
 }
 
 static struct fq_flow *fq_flow_classify(struct fq *fq,
 					struct fq_tin *tin, u32 idx,
-					struct sk_buff *skb,
-					fq_flow_get_default_t get_default_func)
+					struct sk_buff *skb)
 {
 	struct fq_flow *flow;
 
@@ -124,7 +151,7 @@ static struct fq_flow *fq_flow_classify(struct fq *fq,
 
 	flow = &fq->flows[idx];
 	if (flow->tin && flow->tin != tin) {
-		flow = get_default_func(fq, tin, idx, skb);
+		flow = &tin->default_flow;
 		tin->collisions++;
 		fq->collisions++;
 	}
@@ -135,36 +162,56 @@ static struct fq_flow *fq_flow_classify(struct fq *fq,
 	return flow;
 }
 
-static void fq_recalc_backlog(struct fq *fq,
-			      struct fq_tin *tin,
-			      struct fq_flow *flow)
+static struct fq_flow *fq_find_fattest_flow(struct fq *fq)
 {
-	struct fq_flow *i;
+	struct fq_tin *tin;
+	struct fq_flow *flow = NULL;
+	u32 len = 0;
+	int i;
 
-	if (list_empty(&flow->backlogchain))
-		list_add_tail(&flow->backlogchain, &fq->backlogs);
+	for_each_set_bit(i, fq->flows_bitmap, fq->flows_cnt) {
+		struct fq_flow *cur = &fq->flows[i];
+		unsigned int cur_len;
 
-	i = flow;
-	list_for_each_entry_continue_reverse(i, &fq->backlogs,
-					     backlogchain)
-		if (i->backlog > flow->backlog)
-			break;
+		cur_len = cur->backlog;
+		if (cur_len <= len)
+			continue;
+
+		flow = cur;
+		len = cur_len;
+	}
+
+	list_for_each_entry(tin, &fq->tin_backlog, tin_list) {
+		unsigned int cur_len = tin->default_flow.backlog;
 
-	list_move(&flow->backlogchain, &i->backlogchain);
+		if (cur_len <= len)
+			continue;
+
+		flow = &tin->default_flow;
+		len = cur_len;
+	}
+
+	return flow;
 }
 
 static void fq_tin_enqueue(struct fq *fq,
 			   struct fq_tin *tin, u32 idx,
 			   struct sk_buff *skb,
-			   fq_skb_free_t free_func,
-			   fq_flow_get_default_t get_default_func)
+			   fq_skb_free_t free_func)
 {
 	struct fq_flow *flow;
 	bool oom;
 
 	lockdep_assert_held(&fq->lock);
 
-	flow = fq_flow_classify(fq, tin, idx, skb, get_default_func);
+	flow = fq_flow_classify(fq, tin, idx, skb);
+
+	if (!flow->backlog) {
+		if (flow != &tin->default_flow)
+			__set_bit(idx, fq->flows_bitmap);
+		else if (list_empty(&tin->tin_list))
+			list_add(&tin->tin_list, &fq->tin_backlog);
+	}
 
 	flow->tin = tin;
 	flow->backlog += skb->len;
@@ -173,8 +220,6 @@ static void fq_tin_enqueue(struct fq *fq,
 	fq->memory_usage += skb->truesize;
 	fq->backlog++;
 
-	fq_recalc_backlog(fq, tin, flow);
-
 	if (list_empty(&flow->flowchain)) {
 		flow->deficit = fq->quantum;
 		list_add_tail(&flow->flowchain,
@@ -184,18 +229,13 @@ static void fq_tin_enqueue(struct fq *fq,
 	__skb_queue_tail(&flow->queue, skb);
 	oom = (fq->memory_usage > fq->memory_limit);
 	while (fq->backlog > fq->limit || oom) {
-		flow = list_first_entry_or_null(&fq->backlogs,
-						struct fq_flow,
-						backlogchain);
+		flow = fq_find_fattest_flow(fq);
 		if (!flow)
 			return;
 
-		skb = fq_flow_dequeue(fq, flow);
-		if (!skb)
+		if (!fq_flow_drop(fq, flow, free_func))
 			return;
 
-		free_func(fq, flow->tin, flow, skb);
-
 		flow->tin->overlimit++;
 		fq->overlimit++;
 		if (oom) {
@@ -224,8 +264,6 @@ static void fq_flow_filter(struct fq *fq,
 		fq_adjust_removal(fq, flow, skb);
 		free_func(fq, tin, flow, skb);
 	}
-
-	fq_rejigger_backlog(fq, flow);
 }
 
 static void fq_tin_filter(struct fq *fq,
@@ -248,16 +286,18 @@ static void fq_flow_reset(struct fq *fq,
 			  struct fq_flow *flow,
 			  fq_skb_free_t free_func)
 {
+	struct fq_tin *tin = flow->tin;
 	struct sk_buff *skb;
 
 	while ((skb = fq_flow_dequeue(fq, flow)))
-		free_func(fq, flow->tin, flow, skb);
+		free_func(fq, tin, flow, skb);
 
-	if (!list_empty(&flow->flowchain))
+	if (!list_empty(&flow->flowchain)) {
 		list_del_init(&flow->flowchain);
-
-	if (!list_empty(&flow->backlogchain))
-		list_del_init(&flow->backlogchain);
+		if (list_empty(&tin->new_flows) &&
+		    list_empty(&tin->old_flows))
+			list_del_init(&tin->tin_list);
+	}
 
 	flow->tin = NULL;
 
@@ -283,6 +323,7 @@ static void fq_tin_reset(struct fq *fq,
 		fq_flow_reset(fq, flow, free_func);
 	}
 
+	WARN_ON_ONCE(!list_empty(&tin->tin_list));
 	WARN_ON_ONCE(tin->backlog_bytes);
 	WARN_ON_ONCE(tin->backlog_packets);
 }
@@ -290,7 +331,6 @@ static void fq_tin_reset(struct fq *fq,
 static void fq_flow_init(struct fq_flow *flow)
 {
 	INIT_LIST_HEAD(&flow->flowchain);
-	INIT_LIST_HEAD(&flow->backlogchain);
 	__skb_queue_head_init(&flow->queue);
 }
 
@@ -298,6 +338,8 @@ static void fq_tin_init(struct fq_tin *tin)
 {
 	INIT_LIST_HEAD(&tin->new_flows);
 	INIT_LIST_HEAD(&tin->old_flows);
+	INIT_LIST_HEAD(&tin->tin_list);
+	fq_flow_init(&tin->default_flow);
 }
 
 static int fq_init(struct fq *fq, int flows_cnt)
@@ -305,10 +347,9 @@ static int fq_init(struct fq *fq, int flows_cnt)
 	int i;
 
 	memset(fq, 0, sizeof(fq[0]));
-	INIT_LIST_HEAD(&fq->backlogs);
 	spin_lock_init(&fq->lock);
+	INIT_LIST_HEAD(&fq->tin_backlog);
 	fq->flows_cnt = max_t(u32, flows_cnt, 1);
-	get_random_bytes(&fq->perturbation, sizeof(fq->perturbation));
 	fq->quantum = 300;
 	fq->limit = 8192;
 	fq->memory_limit = 16 << 20; /* 16 MBytes */
@@ -317,6 +358,13 @@ static int fq_init(struct fq *fq, int flows_cnt)
 	if (!fq->flows)
 		return -ENOMEM;
 
+	fq->flows_bitmap = bitmap_zalloc(fq->flows_cnt, GFP_KERNEL);
+	if (!fq->flows_bitmap) {
+		kvfree(fq->flows);
+		fq->flows = NULL;
+		return -ENOMEM;
+	}
+
 	for (i = 0; i < fq->flows_cnt; i++)
 		fq_flow_init(&fq->flows[i]);
 
@@ -333,6 +381,9 @@ static void fq_reset(struct fq *fq,
 
 	kvfree(fq->flows);
 	fq->flows = NULL;
+
+	bitmap_free(fq->flows_bitmap);
+	fq->flows_bitmap = NULL;
 }
 
 #endif