[RFC PATCH 08/12] bpf: add devmap, a map for storing net device references

[John Fastabend <john.fastabend@gmail.com>]

Device map (devmap) is a BPF map, primarily useful for networking
applications, that uses a key to lookup a reference to a netdevice.

The map provides a clean way for BPF programs to build virtual port
to physical port maps. Additionally, it provides a scoping function
for the redirect action itself allowing multiple optimizations. Future
patches will leverage the map to provide batching at the XDP layer.

Another optimization/feature, that is not yet implemented, would be
to support multiple netdevices per key to support efficient multicast
and broadcast support.

Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
---
 include/linux/bpf_types.h               |    3 
 include/uapi/linux/bpf.h                |    1 
 kernel/bpf/Makefile                     |    3 
 kernel/bpf/devmap.c                     |  264 +++++++++++++++++++++++++++++++
 kernel/bpf/verifier.c                   |    8 +
 tools/testing/selftests/bpf/test_maps.c |   15 ++
 6 files changed, 294 insertions(+)
 create mode 100644 kernel/bpf/devmap.c

diff --git a/include/linux/bpf_types.h b/include/linux/bpf_types.h
index 03bf223..b18056b 100644
--- a/include/linux/bpf_types.h
+++ b/include/linux/bpf_types.h
@@ -34,3 +34,6 @@
 #endif
 BPF_MAP_TYPE(BPF_MAP_TYPE_ARRAY_OF_MAPS, array_of_maps_map_ops)
 BPF_MAP_TYPE(BPF_MAP_TYPE_HASH_OF_MAPS, htab_of_maps_map_ops)
+#ifdef CONFIG_NET
+BPF_MAP_TYPE(BPF_MAP_TYPE_DEVMAP, dev_map_ops)
+#endif
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index e1f3827..0a48060 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -104,6 +104,7 @@ enum bpf_map_type {
 	BPF_MAP_TYPE_LPM_TRIE,
 	BPF_MAP_TYPE_ARRAY_OF_MAPS,
 	BPF_MAP_TYPE_HASH_OF_MAPS,
+	BPF_MAP_TYPE_DEVMAP,
 };
 
 enum bpf_prog_type {
diff --git a/kernel/bpf/Makefile b/kernel/bpf/Makefile
index e1e5e65..48e9270 100644
--- a/kernel/bpf/Makefile
+++ b/kernel/bpf/Makefile
@@ -2,6 +2,9 @@ obj-y := core.o
 
 obj-$(CONFIG_BPF_SYSCALL) += syscall.o verifier.o inode.o helpers.o
 obj-$(CONFIG_BPF_SYSCALL) += hashtab.o arraymap.o percpu_freelist.o bpf_lru_list.o lpm_trie.o map_in_map.o
+ifeq ($(CONFIG_NET),y)
+obj-$(CONFIG_BPF_SYSCALL) += devmap.o
+endif
 ifeq ($(CONFIG_PERF_EVENTS),y)
 obj-$(CONFIG_BPF_SYSCALL) += stackmap.o
 endif
diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c
new file mode 100644
index 0000000..1a87835
--- /dev/null
+++ b/kernel/bpf/devmap.c
@@ -0,0 +1,264 @@
+/* Copyright (c) 2017 Covalent IO, Inc. http://covalent.io
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of version 2 of the GNU General Public
+ * License as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ */
+
+/* Devmaps primary use is as a backend map for XDP BPF helper call
+ * bpf_redirect_map(). Because XDP is mostly concerned with performance we
+ * spent some effort to ensure the datapath with redirect maps does not use
+ * any locking. This is a quick note on the details.
+ *
+ * We have three possible paths to get into the devmap control plane bpf
+ * syscalls, bpf programs, and driver side xmit/flush operations. A bpf syscall
+ * will invoke an update, delete, or lookup operation. To ensure updates and
+ * deletes appear atomic from the datapath side xchg() is used to modify the
+ * netdev_map array. Then because the datapath does a lookup into the netdev_map
+ * array (read-only) from an RCU critical section we use call_rcu() to wait for
+ * an rcu grace period before free'ing the old data structures. This ensures the
+ * datapath always has a valid copy. However, the datapath does a "flush"
+ * operation that pushes any pending packets in the driver outside the RCU
+ * critical section. Each bpf_dtab_netdev tracks these pending operations using
+ * an atomic per-cpu bitmap. The bpf_dtab_netdev object will not be destroyed
+ * until all bits are cleared indicating outstanding flush operations have
+ * completed.
+ *
+ * BPF syscalls may race with BPF program calls on any of the update, delete
+ * or lookup operations. As noted above the xchg() operation also keep the
+ * netdev_map consistent in this case. From the devmap side BPF programs
+ * calling into these operations are the same as multiple user space threads
+ * making system calls.
+ */
+#include <linux/bpf.h>
+#include <linux/jhash.h>
+#include <linux/filter.h>
+#include <linux/rculist_nulls.h>
+#include "percpu_freelist.h"
+#include "bpf_lru_list.h"
+#include "map_in_map.h"
+
+struct bpf_dtab_netdev {
+	struct net_device *dev;
+	int key;
+	struct rcu_head rcu;
+	struct bpf_dtab *dtab;
+};
+
+struct bpf_dtab {
+	struct bpf_map map;
+	struct bpf_dtab_netdev **netdev_map;
+};
+
+static struct bpf_map *dev_map_alloc(union bpf_attr *attr)
+{
+	struct bpf_dtab *dtab;
+	u64 cost;
+	int err;
+
+	/* check sanity of attributes */
+	if (attr->max_entries == 0 || attr->key_size != 4 ||
+	    attr->value_size != 4 || attr->map_flags)
+		return ERR_PTR(-EINVAL);
+
+	/* if value_size is bigger, the user space won't be able to
+	 * access the elements.
+	 */
+	if (attr->value_size > KMALLOC_MAX_SIZE)
+		return ERR_PTR(-E2BIG);
+
+	dtab = kzalloc(sizeof(*dtab), GFP_USER);
+	if (!dtab)
+		return ERR_PTR(-ENOMEM);
+
+	/* mandatory map attributes */
+	dtab->map.map_type = attr->map_type;
+	dtab->map.key_size = attr->key_size;
+	dtab->map.value_size = attr->value_size;
+	dtab->map.max_entries = attr->max_entries;
+	dtab->map.map_flags = attr->map_flags;
+
+	err = -ENOMEM;
+
+	/* make sure page count doesn't overflow */
+	cost = (u64) dtab->map.max_entries * sizeof(struct bpf_dtab_netdev *);
+	if (cost >= U32_MAX - PAGE_SIZE)
+		goto free_dtab;
+
+	dtab->map.pages = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
+
+	/* if map size is larger than memlock limit, reject it early */
+	err = bpf_map_precharge_memlock(dtab->map.pages);
+	if (err)
+		goto free_dtab;
+
+	dtab->netdev_map = bpf_map_area_alloc(dtab->map.max_entries *
+					      sizeof(struct bpf_dtab_netdev *));
+	if (!dtab->netdev_map)
+		goto free_dtab;
+
+	return &dtab->map;
+
+free_dtab:
+	kfree(dtab);
+	return ERR_PTR(err);
+}
+
+static void dev_map_free(struct bpf_map *map)
+{
+	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
+	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
+	 * disconnected from events. Wait for outstanding critical sections in
+	 * these programs to complete. The rcu critical section only guarantees
+	 * no further reads against netdev_map. It does __not__ ensure pending
+	 * flush operations (if any) are complete.
+	 */
+	synchronize_rcu();
+
+	for (i = 0; i < dtab->map.max_entries; i++) {
+		struct bpf_dtab_netdev *dev;
+
+		dev = dtab->netdev_map[i];
+		if (!dev)
+			continue;
+
+		dev_put(dev->dev);
+		kfree(dev);
+	}
+
+	/* At this point bpf program is detached and all pending operations
+	 * _must_ be complete
+	 */
+	bpf_map_area_free(dtab->netdev_map);
+	kfree(dtab);
+}
+
+static int dev_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
+{
+	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
+	u32 index = key ? *(u32 *)key : U32_MAX;
+	u32 *next = (u32 *)next_key;
+
+	if (index >= dtab->map.max_entries) {
+		*next = 0;
+		return 0;
+	}
+
+	if (index == dtab->map.max_entries - 1)
+		return -ENOENT;
+
+	*next = index + 1;
+	return 0;
+}
+
+/* rcu_read_lock (from syscall and BPF contexts) ensures that if a delete and/or
+ * update happens in parallel here a dev_put wont happen until after reading the
+ * ifindex.
+ */
+static void *dev_map_lookup_elem(struct bpf_map *map, void *key)
+{
+	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
+	struct bpf_dtab_netdev *dev;
+	u32 i = *(u32 *)key;
+
+	if (i >= map->max_entries)
+		return NULL;
+
+	dev = READ_ONCE(dtab->netdev_map[i]);
+	return dev ? &dev->dev->ifindex : NULL;
+}
+
+static void __dev_map_entry_free(struct rcu_head *rcu)
+{
+	struct bpf_dtab_netdev *old_dev;
+
+	old_dev = container_of(rcu, struct bpf_dtab_netdev, rcu);
+	dev_put(old_dev->dev);
+	kfree(old_dev);
+}
+
+static int dev_map_delete_elem(struct bpf_map *map, void *key)
+{
+	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
+	struct bpf_dtab_netdev *old_dev;
+	int k = *(u32 *)key;
+
+	if (k >= map->max_entries)
+		return -EINVAL;
+
+	/* Use synchronize_rcu() here to ensure any rcu critical sections
+	 * have completed, but this does not guarantee a flush has happened
+	 * yet. Because driver side rcu_read_lock/unlock only protects the
+	 * running XDP program. However, for pending flush operations the
+	 * dev and ctx are stored in another per cpu map. And additionally,
+	 * the driver tear down ensures all soft irqs are complete before
+	 * removing the net device in the case of dev_put equals zero.
+	 */
+	old_dev = xchg(&dtab->netdev_map[k], NULL);
+	if (old_dev)
+		call_rcu(&old_dev->rcu, __dev_map_entry_free);
+	return 0;
+}
+
+static int dev_map_update_elem(struct bpf_map *map, void *key, void *value,
+				u64 map_flags)
+{
+	struct bpf_dtab *dtab = container_of(map, struct bpf_dtab, map);
+	struct net *net = current->nsproxy->net_ns;
+	struct bpf_dtab_netdev *dev, *old_dev;
+	u32 i = *(u32 *)key;
+	u32 ifindex = *(u32 *)value;
+
+	if (unlikely(map_flags > BPF_EXIST))
+		return -EINVAL;
+
+	if (unlikely(i >= dtab->map.max_entries))
+		return -E2BIG;
+
+	if (unlikely(map_flags == BPF_NOEXIST))
+		return -EEXIST;
+
+	if (!ifindex) {
+		dev = NULL;
+	} else {
+		dev = kmalloc(sizeof(*dev), GFP_ATOMIC | __GFP_NOWARN);
+		if (!dev)
+			return -ENOMEM;
+
+		dev->dev = dev_get_by_index(net, ifindex);
+		if (!dev->dev) {
+			kfree(dev);
+			return -EINVAL;
+		}
+
+		dev->key = i;
+		dev->dtab = dtab;
+	}
+
+	/* Use call_rcu() here to ensure rcu critical sections have completed
+	 * Remembering the driver side flush operation will happen before the
+	 * net device is removed.
+	 */
+	old_dev = xchg(&dtab->netdev_map[i], dev);
+	if (old_dev)
+		call_rcu(&old_dev->rcu, __dev_map_entry_free);
+
+	return 0;
+}
+
+const struct bpf_map_ops dev_map_ops = {
+	.map_alloc = dev_map_alloc,
+	.map_free = dev_map_free,
+	.map_get_next_key = dev_map_get_next_key,
+	.map_lookup_elem = dev_map_lookup_elem,
+	.map_update_elem = dev_map_update_elem,
+	.map_delete_elem = dev_map_delete_elem,
+};
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 74ea96e..06073ba 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -1294,6 +1294,14 @@ static int check_map_func_compatibility(struct bpf_map *map, int func_id)
 		    func_id != BPF_FUNC_current_task_under_cgroup)
 			goto error;
 		break;
+	/* devmap returns a pointer to a live net_device ifindex that we cannot
+	 * allow to be modified from bpf side. So do not allow lookup elemnts
+	 * for now.
+	 */
+	case BPF_MAP_TYPE_DEVMAP:
+		if (func_id == BPF_FUNC_map_lookup_elem)
+			goto error;
+		break;
 	case BPF_MAP_TYPE_ARRAY_OF_MAPS:
 	case BPF_MAP_TYPE_HASH_OF_MAPS:
 		if (func_id != BPF_FUNC_map_lookup_elem)
diff --git a/tools/testing/selftests/bpf/test_maps.c b/tools/testing/selftests/bpf/test_maps.c
index 79601c8..36d6ac3 100644
--- a/tools/testing/selftests/bpf/test_maps.c
+++ b/tools/testing/selftests/bpf/test_maps.c
@@ -438,6 +438,21 @@ static void test_arraymap_percpu_many_keys(void)
 	close(fd);
 }
 
+static void test_devmap(int task, void *data)
+{
+	int next_key, fd;
+	__u32 key, value;
+
+	fd = bpf_create_map(BPF_MAP_TYPE_DEVMAP, sizeof(key), sizeof(value),
+			    2, 0);
+	if (fd < 0) {
+		printf("Failed to create arraymap '%s'!\n", strerror(errno));
+		exit(1);
+	}
+
+	close(fd);
+}
+
 #define MAP_SIZE (32 * 1024)
 
 static void test_map_large(void)