Merge branch 'simplify-do_redirect'

Björn Töpel says:

====================
This series aims to simplify the XDP maps and
xdp_do_redirect_map()/xdp_do_flush_map(), and to crank out some more
performance from XDP_REDIRECT scenarios.

The first part of the series simplifies all XDP_REDIRECT capable maps,
so that __XXX_flush_map() does not require the map parameter, by
moving the flush list from the map to global scope.

This results in that the map_to_flush member can be removed from
struct bpf_redirect_info, and its corresponding logic.

Simpler code, and more performance due to that checks/code per-packet
is moved to flush.

Pre-series performance:
  $ sudo taskset -c 22 ./xdpsock -i enp134s0f0 -q 20 -n 1 -r -z

   sock0@enp134s0f0:20 rxdrop xdp-drv
                  pps         pkts        1.00
  rx              20,797,350  230,942,399
  tx              0           0

  $ sudo ./xdp_redirect_cpu --dev enp134s0f0 --cpu 22 xdp_cpu_map0

  Running XDP/eBPF prog_name:xdp_cpu_map5_lb_hash_ip_pairs
  XDP-cpumap      CPU:to  pps            drop-pps    extra-info
  XDP-RX          20      7723038        0           0
  XDP-RX          total   7723038        0
  cpumap_kthread  total   0              0           0
  redirect_err    total   0              0
  xdp_exception   total   0              0

Post-series performance:
  $ sudo taskset -c 22 ./xdpsock -i enp134s0f0 -q 20 -n 1 -r -z

   sock0@enp134s0f0:20 rxdrop xdp-drv
                  pps         pkts        1.00
  rx              21,524,979  86,835,327
  tx              0           0

  $ sudo ./xdp_redirect_cpu --dev enp134s0f0 --cpu 22 xdp_cpu_map0

  Running XDP/eBPF prog_name:xdp_cpu_map5_lb_hash_ip_pairs
  XDP-cpumap      CPU:to  pps            drop-pps    extra-info
  XDP-RX          20      7840124        0           0
  XDP-RX          total   7840124        0
  cpumap_kthread  total   0              0           0
  redirect_err    total   0              0
  xdp_exception   total   0              0

Results: +3.5% and +1.5% for the ubenchmarks.

v1->v2 [1]:
  * Removed 'unused-variable' compiler warning (Jakub)

[1] https://lore.kernel.org/bpf/20191218105400.2895-1-bjorn.topel@gmail.com/
====================

Signed-off-by: Alexei Starovoitov <ast@kernel.org>

3 files changed, 47 insertions, 125 deletions

diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index ef49e17ae47c..70f71b154fa5 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c
@@ -72,17 +72,18 @@ struct bpf_cpu_map {
 	struct bpf_map map;
 	/* Below members specific for map type */
 	struct bpf_cpu_map_entry **cpu_map;
-	struct list_head __percpu *flush_list;
 };
 
-static int bq_flush_to_queue(struct xdp_bulk_queue *bq, bool in_napi_ctx);
+static DEFINE_PER_CPU(struct list_head, cpu_map_flush_list);
+
+static int bq_flush_to_queue(struct xdp_bulk_queue *bq);
 
 static struct bpf_map *cpu_map_alloc(union bpf_attr *attr)
 {
 	struct bpf_cpu_map *cmap;
 	int err = -ENOMEM;
-	int ret, cpu;
 	u64 cost;
+	int ret;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return ERR_PTR(-EPERM);
@@ -106,7 +107,6 @@ static struct bpf_map *cpu_map_alloc(union bpf_attr *attr)
 
 	/* make sure page count doesn't overflow */
 	cost = (u64) cmap->map.max_entries * sizeof(struct bpf_cpu_map_entry *);
-	cost += sizeof(struct list_head) * num_possible_cpus();
 
 	/* Notice returns -EPERM on if map size is larger than memlock limit */
 	ret = bpf_map_charge_init(&cmap->map.memory, cost);
@@ -115,23 +115,14 @@ static struct bpf_map *cpu_map_alloc(union bpf_attr *attr)
 		goto free_cmap;
 	}
 
-	cmap->flush_list = alloc_percpu(struct list_head);
-	if (!cmap->flush_list)
-		goto free_charge;
-
-	for_each_possible_cpu(cpu)
-		INIT_LIST_HEAD(per_cpu_ptr(cmap->flush_list, cpu));
-
 	/* Alloc array for possible remote "destination" CPUs */
 	cmap->cpu_map = bpf_map_area_alloc(cmap->map.max_entries *
 					   sizeof(struct bpf_cpu_map_entry *),
 					   cmap->map.numa_node);
 	if (!cmap->cpu_map)
-		goto free_percpu;
+		goto free_charge;
 
 	return &cmap->map;
-free_percpu:
-	free_percpu(cmap->flush_list);
 free_charge:
 	bpf_map_charge_finish(&cmap->map.memory);
 free_cmap:
@@ -399,22 +390,14 @@ free_rcu:
 static void __cpu_map_entry_free(struct rcu_head *rcu)
 {
 	struct bpf_cpu_map_entry *rcpu;
-	int cpu;
 
 	/* This cpu_map_entry have been disconnected from map and one
-	 * RCU graze-period have elapsed.  Thus, XDP cannot queue any
+	 * RCU grace-period have elapsed.  Thus, XDP cannot queue any
 	 * new packets and cannot change/set flush_needed that can
 	 * find this entry.
 	 */
 	rcpu = container_of(rcu, struct bpf_cpu_map_entry, rcu);
 
-	/* Flush remaining packets in percpu bulkq */
-	for_each_online_cpu(cpu) {
-		struct xdp_bulk_queue *bq = per_cpu_ptr(rcpu->bulkq, cpu);
-
-		/* No concurrent bq_enqueue can run at this point */
-		bq_flush_to_queue(bq, false);
-	}
 	free_percpu(rcpu->bulkq);
 	/* Cannot kthread_stop() here, last put free rcpu resources */
 	put_cpu_map_entry(rcpu);
@@ -436,7 +419,7 @@ static void __cpu_map_entry_free(struct rcu_head *rcu)
  * percpu bulkq to queue.  Due to caller map_delete_elem() disable
  * preemption, cannot call kthread_stop() to make sure queue is empty.
  * Instead a work_queue is started for stopping kthread,
- * cpu_map_kthread_stop, which waits for an RCU graze period before
+ * cpu_map_kthread_stop, which waits for an RCU grace period before
  * stopping kthread, emptying the queue.
  */
 static void __cpu_map_entry_replace(struct bpf_cpu_map *cmap,
@@ -507,7 +490,6 @@ static int cpu_map_update_elem(struct bpf_map *map, void *key, void *value,
 static void cpu_map_free(struct bpf_map *map)
 {
 	struct bpf_cpu_map *cmap = container_of(map, struct bpf_cpu_map, map);
-	int cpu;
 	u32 i;
 
 	/* At this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
@@ -522,18 +504,6 @@ static void cpu_map_free(struct bpf_map *map)
 	bpf_clear_redirect_map(map);
 	synchronize_rcu();
 
-	/* To ensure all pending flush operations have completed wait for flush
-	 * list be empty on _all_ cpus. Because the above synchronize_rcu()
-	 * ensures the map is disconnected from the program we can assume no new
-	 * items will be added to the list.
-	 */
-	for_each_online_cpu(cpu) {
-		struct list_head *flush_list = per_cpu_ptr(cmap->flush_list, cpu);
-
-		while (!list_empty(flush_list))
-			cond_resched();
-	}
-
 	/* For cpu_map the remote CPUs can still be using the entries
 	 * (struct bpf_cpu_map_entry).
 	 */
@@ -544,10 +514,9 @@ static void cpu_map_free(struct bpf_map *map)
 		if (!rcpu)
 			continue;
 
-		/* bq flush and cleanup happens after RCU graze-period */
+		/* bq flush and cleanup happens after RCU grace-period */
 		__cpu_map_entry_replace(cmap, i, NULL); /* call_rcu */
 	}
-	free_percpu(cmap->flush_list);
 	bpf_map_area_free(cmap->cpu_map);
 	kfree(cmap);
 }
@@ -599,7 +568,7 @@ const struct bpf_map_ops cpu_map_ops = {
 	.map_check_btf		= map_check_no_btf,
 };
 
-static int bq_flush_to_queue(struct xdp_bulk_queue *bq, bool in_napi_ctx)
+static int bq_flush_to_queue(struct xdp_bulk_queue *bq)
 {
 	struct bpf_cpu_map_entry *rcpu = bq->obj;
 	unsigned int processed = 0, drops = 0;
@@ -620,10 +589,7 @@ static int bq_flush_to_queue(struct xdp_bulk_queue *bq, bool in_napi_ctx)
 		err = __ptr_ring_produce(q, xdpf);
 		if (err) {
 			drops++;
-			if (likely(in_napi_ctx))
-				xdp_return_frame_rx_napi(xdpf);
-			else
-				xdp_return_frame(xdpf);
+			xdp_return_frame_rx_napi(xdpf);
 		}
 		processed++;
 	}
@@ -642,11 +608,11 @@ static int bq_flush_to_queue(struct xdp_bulk_queue *bq, bool in_napi_ctx)
  */
 static int bq_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf)
 {
-	struct list_head *flush_list = this_cpu_ptr(rcpu->cmap->flush_list);
+	struct list_head *flush_list = this_cpu_ptr(&cpu_map_flush_list);
 	struct xdp_bulk_queue *bq = this_cpu_ptr(rcpu->bulkq);
 
 	if (unlikely(bq->count == CPU_MAP_BULK_SIZE))
-		bq_flush_to_queue(bq, true);
+		bq_flush_to_queue(bq);
 
 	/* Notice, xdp_buff/page MUST be queued here, long enough for
 	 * driver to code invoking us to finished, due to driver
@@ -681,16 +647,26 @@ int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
 	return 0;
 }
 
-void __cpu_map_flush(struct bpf_map *map)
+void __cpu_map_flush(void)
 {
-	struct bpf_cpu_map *cmap = container_of(map, struct bpf_cpu_map, map);
-	struct list_head *flush_list = this_cpu_ptr(cmap->flush_list);
+	struct list_head *flush_list = this_cpu_ptr(&cpu_map_flush_list);
 	struct xdp_bulk_queue *bq, *tmp;
 
 	list_for_each_entry_safe(bq, tmp, flush_list, flush_node) {
-		bq_flush_to_queue(bq, true);
+		bq_flush_to_queue(bq);
 
 		/* If already running, costs spin_lock_irqsave + smb_mb */
 		wake_up_process(bq->obj->kthread);
 	}
 }
+
+static int __init cpu_map_init(void)
+{
+	int cpu;
+
+	for_each_possible_cpu(cpu)
+		INIT_LIST_HEAD(&per_cpu(cpu_map_flush_list, cpu));
+	return 0;
+}
+
+subsys_initcall(cpu_map_init);
diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c
index 3d3d61b5985b..da9c832fc5c8 100644
--- a/kernel/bpf/devmap.c
+++ b/kernel/bpf/devmap.c
@@ -75,7 +75,6 @@ struct bpf_dtab_netdev {
 struct bpf_dtab {
 	struct bpf_map map;
 	struct bpf_dtab_netdev **netdev_map; /* DEVMAP type only */
-	struct list_head __percpu *flush_list;
 	struct list_head list;
 
 	/* these are only used for DEVMAP_HASH type maps */
@@ -85,6 +84,7 @@ struct bpf_dtab {
 	u32 n_buckets;
 };
 
+static DEFINE_PER_CPU(struct list_head, dev_map_flush_list);
 static DEFINE_SPINLOCK(dev_map_lock);
 static LIST_HEAD(dev_map_list);
 
@@ -109,8 +109,8 @@ static inline struct hlist_head *dev_map_index_hash(struct bpf_dtab *dtab,
 
 static int dev_map_init_map(struct bpf_dtab *dtab, union bpf_attr *attr)
 {
-	int err, cpu;
-	u64 cost;
+	u64 cost = 0;
+	int err;
 
 	/* check sanity of attributes */
 	if (attr->max_entries == 0 || attr->key_size != 4 ||
@@ -125,9 +125,6 @@ static int dev_map_init_map(struct bpf_dtab *dtab, union bpf_attr *attr)
 
 	bpf_map_init_from_attr(&dtab->map, attr);
 
-	/* make sure page count doesn't overflow */
-	cost = (u64) sizeof(struct list_head) * num_possible_cpus();
-
 	if (attr->map_type == BPF_MAP_TYPE_DEVMAP_HASH) {
 		dtab->n_buckets = roundup_pow_of_two(dtab->map.max_entries);
 
@@ -143,17 +140,10 @@ static int dev_map_init_map(struct bpf_dtab *dtab, union bpf_attr *attr)
 	if (err)
 		return -EINVAL;
 
-	dtab->flush_list = alloc_percpu(struct list_head);
-	if (!dtab->flush_list)
-		goto free_charge;
-
-	for_each_possible_cpu(cpu)
-		INIT_LIST_HEAD(per_cpu_ptr(dtab->flush_list, cpu));
-
 	if (attr->map_type == BPF_MAP_TYPE_DEVMAP_HASH) {
 		dtab->dev_index_head = dev_map_create_hash(dtab->n_buckets);
 		if (!dtab->dev_index_head)
-			goto free_percpu;
+			goto free_charge;
 
 		spin_lock_init(&dtab->index_lock);
 	} else {
@@ -161,13 +151,11 @@ static int dev_map_init_map(struct bpf_dtab *dtab, union bpf_attr *attr)
 						      sizeof(struct bpf_dtab_netdev *),
 						      dtab->map.numa_node);
 		if (!dtab->netdev_map)
-			goto free_percpu;
+			goto free_charge;
 	}
 
 	return 0;
 
-free_percpu:
-	free_percpu(dtab->flush_list);
 free_charge:
 	bpf_map_charge_finish(&dtab->map.memory);
 	return -ENOMEM;
@@ -201,7 +189,7 @@ static struct bpf_map *dev_map_alloc(union bpf_attr *attr)
 static void dev_map_free(struct bpf_map *map)
 {
 	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
-	int i, cpu;
+	int i;
 
 	/* At this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
 	 * so the programs (can be more than one that used this map) were
@@ -221,18 +209,6 @@ static void dev_map_free(struct bpf_map *map)
 	/* Make sure prior __dev_map_entry_free() have completed. */
 	rcu_barrier();
 
-	/* To ensure all pending flush operations have completed wait for flush
-	 * list to empty on _all_ cpus.
-	 * Because the above synchronize_rcu() ensures the map is disconnected
-	 * from the program we can assume no new items will be added.
-	 */
-	for_each_online_cpu(cpu) {
-		struct list_head *flush_list = per_cpu_ptr(dtab->flush_list, cpu);
-
-		while (!list_empty(flush_list))
-			cond_resched();
-	}
-
 	if (dtab->map.map_type == BPF_MAP_TYPE_DEVMAP_HASH) {
 		for (i = 0; i < dtab->n_buckets; i++) {
 			struct bpf_dtab_netdev *dev;
@@ -266,7 +242,6 @@ static void dev_map_free(struct bpf_map *map)
 		bpf_map_area_free(dtab->netdev_map);
 	}
 
-	free_percpu(dtab->flush_list);
 	kfree(dtab);
 }
 
@@ -345,8 +320,7 @@ static int dev_map_hash_get_next_key(struct bpf_map *map, void *key,
 	return -ENOENT;
 }
 
-static int bq_xmit_all(struct xdp_bulk_queue *bq, u32 flags,
-		       bool in_napi_ctx)
+static int bq_xmit_all(struct xdp_bulk_queue *bq, u32 flags)
 {
 	struct bpf_dtab_netdev *obj = bq->obj;
 	struct net_device *dev = obj->dev;
@@ -384,11 +358,7 @@ error:
 	for (i = 0; i < bq->count; i++) {
 		struct xdp_frame *xdpf = bq->q[i];
 
-		/* RX path under NAPI protection, can return frames faster */
-		if (likely(in_napi_ctx))
-			xdp_return_frame_rx_napi(xdpf);
-		else
-			xdp_return_frame(xdpf);
+		xdp_return_frame_rx_napi(xdpf);
 		drops++;
 	}
 	goto out;
@@ -401,15 +371,14 @@ error:
  * net device can be torn down. On devmap tear down we ensure the flush list
  * is empty before completing to ensure all flush operations have completed.
  */
-void __dev_map_flush(struct bpf_map *map)
+void __dev_map_flush(void)
 {
-	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
-	struct list_head *flush_list = this_cpu_ptr(dtab->flush_list);
+	struct list_head *flush_list = this_cpu_ptr(&dev_map_flush_list);
 	struct xdp_bulk_queue *bq, *tmp;
 
 	rcu_read_lock();
 	list_for_each_entry_safe(bq, tmp, flush_list, flush_node)
-		bq_xmit_all(bq, XDP_XMIT_FLUSH, true);
+		bq_xmit_all(bq, XDP_XMIT_FLUSH);
 	rcu_read_unlock();
 }
 
@@ -436,11 +405,11 @@ static int bq_enqueue(struct bpf_dtab_netdev *obj, struct xdp_frame *xdpf,
 		      struct net_device *dev_rx)
 
 {
-	struct list_head *flush_list = this_cpu_ptr(obj->dtab->flush_list);
+	struct list_head *flush_list = this_cpu_ptr(&dev_map_flush_list);
 	struct xdp_bulk_queue *bq = this_cpu_ptr(obj->bulkq);
 
 	if (unlikely(bq->count == DEV_MAP_BULK_SIZE))
-		bq_xmit_all(bq, 0, true);
+		bq_xmit_all(bq, 0);
 
 	/* Ingress dev_rx will be the same for all xdp_frame's in
 	 * bulk_queue, because bq stored per-CPU and must be flushed
@@ -509,27 +478,11 @@ static void *dev_map_hash_lookup_elem(struct bpf_map *map, void *key)
 	return dev ? &dev->ifindex : NULL;
 }
 
-static void dev_map_flush_old(struct bpf_dtab_netdev *dev)
-{
-	if (dev->dev->netdev_ops->ndo_xdp_xmit) {
-		struct xdp_bulk_queue *bq;
-		int cpu;
-
-		rcu_read_lock();
-		for_each_online_cpu(cpu) {
-			bq = per_cpu_ptr(dev->bulkq, cpu);
-			bq_xmit_all(bq, XDP_XMIT_FLUSH, false);
-		}
-		rcu_read_unlock();
-	}
-}
-
 static void __dev_map_entry_free(struct rcu_head *rcu)
 {
 	struct bpf_dtab_netdev *dev;
 
 	dev = container_of(rcu, struct bpf_dtab_netdev, rcu);
-	dev_map_flush_old(dev);
 	free_percpu(dev->bulkq);
 	dev_put(dev->dev);
 	kfree(dev);
@@ -810,10 +763,15 @@ static struct notifier_block dev_map_notifier = {
 
 static int __init dev_map_init(void)
 {
+	int cpu;
+
 	/* Assure tracepoint shadow struct _bpf_dtab_netdev is in sync */
 	BUILD_BUG_ON(offsetof(struct bpf_dtab_netdev, dev) !=
 		     offsetof(struct _bpf_dtab_netdev, dev));
 	register_netdevice_notifier(&dev_map_notifier);
+
+	for_each_possible_cpu(cpu)
+		INIT_LIST_HEAD(&per_cpu(dev_map_flush_list, cpu));
 	return 0;
 }
 
diff --git a/kernel/bpf/xskmap.c b/kernel/bpf/xskmap.c
index 90c4fce1c981..2cc5c8f4c800 100644
--- a/kernel/bpf/xskmap.c
+++ b/kernel/bpf/xskmap.c
@@ -72,9 +72,9 @@ static void xsk_map_sock_delete(struct xdp_sock *xs,
 static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
 {
 	struct bpf_map_memory mem;
-	int cpu, err, numa_node;
+	int err, numa_node;
 	struct xsk_map *m;
-	u64 cost, size;
+	u64 size;
 
 	if (!capable(CAP_NET_ADMIN))
 		return ERR_PTR(-EPERM);
@@ -86,9 +86,8 @@ static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
 
 	numa_node = bpf_map_attr_numa_node(attr);
 	size = struct_size(m, xsk_map, attr->max_entries);
-	cost = size + array_size(sizeof(*m->flush_list), num_possible_cpus());
 
-	err = bpf_map_charge_init(&mem, cost);
+	err = bpf_map_charge_init(&mem, size);
 	if (err < 0)
 		return ERR_PTR(err);
 
@@ -102,16 +101,6 @@ static struct bpf_map *xsk_map_alloc(union bpf_attr *attr)
 	bpf_map_charge_move(&m->map.memory, &mem);
 	spin_lock_init(&m->lock);
 
-	m->flush_list = alloc_percpu(struct list_head);
-	if (!m->flush_list) {
-		bpf_map_charge_finish(&m->map.memory);
-		bpf_map_area_free(m);
-		return ERR_PTR(-ENOMEM);
-	}
-
-	for_each_possible_cpu(cpu)
-		INIT_LIST_HEAD(per_cpu_ptr(m->flush_list, cpu));
-
 	return &m->map;
 }
 
@@ -121,7 +110,6 @@ static void xsk_map_free(struct bpf_map *map)
 
 	bpf_clear_redirect_map(map);
 	synchronize_net();
-	free_percpu(m->flush_list);
 	bpf_map_area_free(m);
 }