bpf: Introduce BPF_MAP_TYPE_REUSEPORT_SOCKARRAY

This patch introduces a new map type BPF_MAP_TYPE_REUSEPORT_SOCKARRAY.

To unleash the full potential of a bpf prog, it is essential for the
userspace to be capable of directly setting up a bpf map which can then
be consumed by the bpf prog to make decision.  In this case, decide which
SO_REUSEPORT sk to serve the incoming request.

By adding BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, the userspace has total control
and visibility on where a SO_REUSEPORT sk should be located in a bpf map.
The later patch will introduce BPF_PROG_TYPE_SK_REUSEPORT such that
the bpf prog can directly select a sk from the bpf map.  That will
raise the programmability of the bpf prog attached to a reuseport
group (a group of sk serving the same IP:PORT).

For example, in UDP, the bpf prog can peek into the payload (e.g.
through the "data" pointer introduced in the later patch) to learn
the application level's connection information and then decide which sk
to pick from a bpf map.  The userspace can tightly couple the sk's location
in a bpf map with the application logic in generating the UDP payload's
connection information.  This connection info contact/API stays within the
userspace.

Also, when used with map-in-map, the userspace can switch the
old-server-process's inner map to a new-server-process's inner map
in one call "bpf_map_update_elem(outer_map, &index, &new_reuseport_array)".
The bpf prog will then direct incoming requests to the new process instead
of the old process.  The old process can finish draining the pending
requests (e.g. by "accept()") before closing the old-fds.  [Note that
deleting a fd from a bpf map does not necessary mean the fd is closed]

During map_update_elem(),
Only SO_REUSEPORT sk (i.e. which has already been added
to a reuse->socks[]) can be used.  That means a SO_REUSEPORT sk that is
"bind()" for UDP or "bind()+listen()" for TCP.  These conditions are
ensured in "reuseport_array_update_check()".

A SO_REUSEPORT sk can only be added once to a map (i.e. the
same sk cannot be added twice even to the same map).  SO_REUSEPORT
already allows another sk to be created for the same IP:PORT.
There is no need to re-create a similar usage in the BPF side.

When a SO_REUSEPORT is deleted from the "reuse->socks[]" (e.g. "close()"),
it will notify the bpf map to remove it from the map also.  It is
done through "bpf_sk_reuseport_detach()" and it will only be called
if >=1 of the "reuse->sock[]" has ever been added to a bpf map.

The map_update()/map_delete() has to be in-sync with the
"reuse->socks[]".  Hence, the same "reuseport_lock" used
by "reuse->socks[]" has to be used here also. Care has
been taken to ensure the lock is only acquired when the
adding sk passes some strict tests. and
freeing the map does not require the reuseport_lock.

The reuseport_array will also support lookup from the syscall
side.  It will return a sock_gen_cookie().  The sock_gen_cookie()
is on-demand (i.e. a sk's cookie is not generated until the very
first map_lookup_elem()).

The lookup cookie is 64bits but it goes against the logical userspace
expectation on 32bits sizeof(fd) (and as other fd based bpf maps do also).
It may catch user in surprise if we enforce value_size=8 while
userspace still pass a 32bits fd during update.  Supporting different
value_size between lookup and update seems unintuitive also.

We also need to consider what if other existing fd based maps want
to return 64bits value from syscall's lookup in the future.
Hence, reuseport_array supports both value_size 4 and 8, and
assuming user will usually use value_size=4.  The syscall's lookup
will return ENOSPC on value_size=4.  It will will only
return 64bits value from sock_gen_cookie() when user consciously
choose value_size=8 (as a signal that lookup is desired) which then
requires a 64bits value in both lookup and update.

Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

include/linux/bpf.h

@@ -524,6 +524,7 @@ static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
 }
 
 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
+int array_map_alloc_check(union bpf_attr *attr);
 
 #else /* !CONFIG_BPF_SYSCALL */
 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
@@ -769,6 +770,33 @@ static inline void __xsk_map_flush(struct bpf_map *map)
 }
 #endif
 
+#if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
+void bpf_sk_reuseport_detach(struct sock *sk);
+int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
+				       void *value);
+int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
+				       void *value, u64 map_flags);
+#else
+static inline void bpf_sk_reuseport_detach(struct sock *sk)
+{
+}
+
+#ifdef CONFIG_BPF_SYSCALL
+static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
+						     void *key, void *value)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
+						     void *key, void *value,
+						     u64 map_flags)
+{
+	return -EOPNOTSUPP;
+}
+#endif /* CONFIG_BPF_SYSCALL */
+#endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
+
 /* verifier prototypes for helper functions called from eBPF programs */
 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
 extern const struct bpf_func_proto bpf_map_update_elem_proto;

include/linux/bpf_types.h

@@ -60,4 +60,7 @@ BPF_MAP_TYPE(BPF_MAP_TYPE_CPUMAP, cpu_map_ops)
 #if defined(CONFIG_XDP_SOCKETS)
 BPF_MAP_TYPE(BPF_MAP_TYPE_XSKMAP, xsk_map_ops)
 #endif
+#ifdef CONFIG_INET
+BPF_MAP_TYPE(BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, reuseport_array_ops)
+#endif
 #endif

include/uapi/linux/bpf.h

@@ -126,6 +126,7 @@ enum bpf_map_type {
 	BPF_MAP_TYPE_XSKMAP,
 	BPF_MAP_TYPE_SOCKHASH,
 	BPF_MAP_TYPE_CGROUP_STORAGE,
+	BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
 };
 
 enum bpf_prog_type {

kernel/bpf/Makefile

@@ -23,3 +23,6 @@ ifeq ($(CONFIG_PERF_EVENTS),y)
 obj-$(CONFIG_BPF_SYSCALL) += stackmap.o
 endif
 obj-$(CONFIG_CGROUP_BPF) += cgroup.o
+ifeq ($(CONFIG_INET),y)
+obj-$(CONFIG_BPF_SYSCALL) += reuseport_array.o
+endif

kernel/bpf/arraymap.c

@@ -54,7 +54,7 @@ static int bpf_array_alloc_percpu(struct bpf_array *array)
 }
 
 /* Called from syscall */
-static int array_map_alloc_check(union bpf_attr *attr)
+int array_map_alloc_check(union bpf_attr *attr)
 {
 	bool percpu = attr->map_type == BPF_MAP_TYPE_PERCPU_ARRAY;
 	int numa_node = bpf_map_attr_numa_node(attr);

kernel/bpf/reuseport_array.c

@@ -0,0 +1,363 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2018 Facebook
+ */
+#include <linux/bpf.h>
+#include <linux/err.h>
+#include <linux/sock_diag.h>
+#include <net/sock_reuseport.h>
+
+struct reuseport_array {
+	struct bpf_map map;
+	struct sock __rcu *ptrs[];
+};
+
+static struct reuseport_array *reuseport_array(struct bpf_map *map)
+{
+	return (struct reuseport_array *)map;
+}
+
+/* The caller must hold the reuseport_lock */
+void bpf_sk_reuseport_detach(struct sock *sk)
+{
+	struct sock __rcu **socks;
+
+	write_lock_bh(&sk->sk_callback_lock);
+	socks = sk->sk_user_data;
+	if (socks) {
+		WRITE_ONCE(sk->sk_user_data, NULL);
+		/*
+		 * Do not move this NULL assignment outside of
+		 * sk->sk_callback_lock because there is
+		 * a race with reuseport_array_free()
+		 * which does not hold the reuseport_lock.
+		 */
+		RCU_INIT_POINTER(*socks, NULL);
+	}
+	write_unlock_bh(&sk->sk_callback_lock);
+}
+
+static int reuseport_array_alloc_check(union bpf_attr *attr)
+{
+	if (attr->value_size != sizeof(u32) &&
+	    attr->value_size != sizeof(u64))
+		return -EINVAL;
+
+	return array_map_alloc_check(attr);
+}
+
+static void *reuseport_array_lookup_elem(struct bpf_map *map, void *key)
+{
+	struct reuseport_array *array = reuseport_array(map);
+	u32 index = *(u32 *)key;
+
+	if (unlikely(index >= array->map.max_entries))
+		return NULL;
+
+	return rcu_dereference(array->ptrs[index]);
+}
+
+/* Called from syscall only */
+static int reuseport_array_delete_elem(struct bpf_map *map, void *key)
+{
+	struct reuseport_array *array = reuseport_array(map);
+	u32 index = *(u32 *)key;
+	struct sock *sk;
+	int err;
+
+	if (index >= map->max_entries)
+		return -E2BIG;
+
+	if (!rcu_access_pointer(array->ptrs[index]))
+		return -ENOENT;
+
+	spin_lock_bh(&reuseport_lock);
+
+	sk = rcu_dereference_protected(array->ptrs[index],
+				       lockdep_is_held(&reuseport_lock));
+	if (sk) {
+		write_lock_bh(&sk->sk_callback_lock);
+		WRITE_ONCE(sk->sk_user_data, NULL);
+		RCU_INIT_POINTER(array->ptrs[index], NULL);
+		write_unlock_bh(&sk->sk_callback_lock);
+		err = 0;
+	} else {
+		err = -ENOENT;
+	}
+
+	spin_unlock_bh(&reuseport_lock);
+
+	return err;
+}
+
+static void reuseport_array_free(struct bpf_map *map)
+{
+	struct reuseport_array *array = reuseport_array(map);
+	struct sock *sk;
+	u32 i;
+
+	synchronize_rcu();
+
+	/*
+	 * ops->map_*_elem() will not be able to access this
+	 * array now. Hence, this function only races with
+	 * bpf_sk_reuseport_detach() which was triggerred by
+	 * close() or disconnect().
+	 *
+	 * This function and bpf_sk_reuseport_detach() are
+	 * both removing sk from "array".  Who removes it
+	 * first does not matter.
+	 *
+	 * The only concern here is bpf_sk_reuseport_detach()
+	 * may access "array" which is being freed here.
+	 * bpf_sk_reuseport_detach() access this "array"
+	 * through sk->sk_user_data _and_ with sk->sk_callback_lock
+	 * held which is enough because this "array" is not freed
+	 * until all sk->sk_user_data has stopped referencing this "array".
+	 *
+	 * Hence, due to the above, taking "reuseport_lock" is not
+	 * needed here.
+	 */
+
+	/*
+	 * Since reuseport_lock is not taken, sk is accessed under
+	 * rcu_read_lock()
+	 */
+	rcu_read_lock();
+	for (i = 0; i < map->max_entries; i++) {
+		sk = rcu_dereference(array->ptrs[i]);
+		if (sk) {
+			write_lock_bh(&sk->sk_callback_lock);
+			/*
+			 * No need for WRITE_ONCE(). At this point,
+			 * no one is reading it without taking the
+			 * sk->sk_callback_lock.
+			 */
+			sk->sk_user_data = NULL;
+			write_unlock_bh(&sk->sk_callback_lock);
+			RCU_INIT_POINTER(array->ptrs[i], NULL);
+		}
+	}
+	rcu_read_unlock();
+
+	/*
+	 * Once reaching here, all sk->sk_user_data is not
+	 * referenceing this "array".  "array" can be freed now.
+	 */
+	bpf_map_area_free(array);
+}
+
+static struct bpf_map *reuseport_array_alloc(union bpf_attr *attr)
+{
+	int err, numa_node = bpf_map_attr_numa_node(attr);
+	struct reuseport_array *array;
+	u64 cost, array_size;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return ERR_PTR(-EPERM);
+
+	array_size = sizeof(*array);
+	array_size += (u64)attr->max_entries * sizeof(struct sock *);
+
+	/* make sure there is no u32 overflow later in round_up() */
+	cost = array_size;
+	if (cost >= U32_MAX - PAGE_SIZE)
+		return ERR_PTR(-ENOMEM);
+	cost = round_up(cost, PAGE_SIZE) >> PAGE_SHIFT;
+
+	err = bpf_map_precharge_memlock(cost);
+	if (err)
+		return ERR_PTR(err);
+
+	/* allocate all map elements and zero-initialize them */
+	array = bpf_map_area_alloc(array_size, numa_node);
+	if (!array)
+		return ERR_PTR(-ENOMEM);
+
+	/* copy mandatory map attributes */
+	bpf_map_init_from_attr(&array->map, attr);
+	array->map.pages = cost;
+
+	return &array->map;
+}
+
+int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
+				       void *value)
+{
+	struct sock *sk;
+	int err;
+
+	if (map->value_size != sizeof(u64))
+		return -ENOSPC;
+
+	rcu_read_lock();
+	sk = reuseport_array_lookup_elem(map, key);
+	if (sk) {
+		*(u64 *)value = sock_gen_cookie(sk);
+		err = 0;
+	} else {
+		err = -ENOENT;
+	}
+	rcu_read_unlock();
+
+	return err;
+}
+
+static int
+reuseport_array_update_check(const struct reuseport_array *array,
+			     const struct sock *nsk,
+			     const struct sock *osk,
+			     const struct sock_reuseport *nsk_reuse,
+			     u32 map_flags)
+{
+	if (osk && map_flags == BPF_NOEXIST)
+		return -EEXIST;
+
+	if (!osk && map_flags == BPF_EXIST)
+		return -ENOENT;
+
+	if (nsk->sk_protocol != IPPROTO_UDP && nsk->sk_protocol != IPPROTO_TCP)
+		return -ENOTSUPP;
+
+	if (nsk->sk_family != AF_INET && nsk->sk_family != AF_INET6)
+		return -ENOTSUPP;
+
+	if (nsk->sk_type != SOCK_STREAM && nsk->sk_type != SOCK_DGRAM)
+		return -ENOTSUPP;
+
+	/*
+	 * sk must be hashed (i.e. listening in the TCP case or binded
+	 * in the UDP case) and
+	 * it must also be a SO_REUSEPORT sk (i.e. reuse cannot be NULL).
+	 *
+	 * Also, sk will be used in bpf helper that is protected by
+	 * rcu_read_lock().
+	 */
+	if (!sock_flag(nsk, SOCK_RCU_FREE) || !sk_hashed(nsk) || !nsk_reuse)
+		return -EINVAL;
+
+	/* READ_ONCE because the sk->sk_callback_lock may not be held here */
+	if (READ_ONCE(nsk->sk_user_data))
+		return -EBUSY;
+
+	return 0;
+}
+
+/*
+ * Called from syscall only.
+ * The "nsk" in the fd refcnt.
+ * The "osk" and "reuse" are protected by reuseport_lock.
+ */
+int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
+				       void *value, u64 map_flags)
+{
+	struct reuseport_array *array = reuseport_array(map);
+	struct sock *free_osk = NULL, *osk, *nsk;
+	struct sock_reuseport *reuse;
+	u32 index = *(u32 *)key;
+	struct socket *socket;
+	int err, fd;
+
+	if (map_flags > BPF_EXIST)
+		return -EINVAL;
+
+	if (index >= map->max_entries)
+		return -E2BIG;
+
+	if (map->value_size == sizeof(u64)) {
+		u64 fd64 = *(u64 *)value;
+
+		if (fd64 > S32_MAX)
+			return -EINVAL;
+		fd = fd64;
+	} else {
+		fd = *(int *)value;
+	}
+
+	socket = sockfd_lookup(fd, &err);
+	if (!socket)
+		return err;
+
+	nsk = socket->sk;
+	if (!nsk) {
+		err = -EINVAL;
+		goto put_file;
+	}
+
+	/* Quick checks before taking reuseport_lock */
+	err = reuseport_array_update_check(array, nsk,
+					   rcu_access_pointer(array->ptrs[index]),
+					   rcu_access_pointer(nsk->sk_reuseport_cb),
+					   map_flags);
+	if (err)
+		goto put_file;
+
+	spin_lock_bh(&reuseport_lock);
+	/*
+	 * Some of the checks only need reuseport_lock
+	 * but it is done under sk_callback_lock also
+	 * for simplicity reason.
+	 */
+	write_lock_bh(&nsk->sk_callback_lock);
+
+	osk = rcu_dereference_protected(array->ptrs[index],
+					lockdep_is_held(&reuseport_lock));
+	reuse = rcu_dereference_protected(nsk->sk_reuseport_cb,
+					  lockdep_is_held(&reuseport_lock));
+	err = reuseport_array_update_check(array, nsk, osk, reuse, map_flags);
+	if (err)
+		goto put_file_unlock;
+
+	/* Ensure reuse->reuseport_id is set */
+	err = reuseport_get_id(reuse);
+	if (err < 0)
+		goto put_file_unlock;
+
+	WRITE_ONCE(nsk->sk_user_data, &array->ptrs[index]);
+	rcu_assign_pointer(array->ptrs[index], nsk);
+	free_osk = osk;
+	err = 0;
+
+put_file_unlock:
+	write_unlock_bh(&nsk->sk_callback_lock);
+
+	if (free_osk) {
+		write_lock_bh(&free_osk->sk_callback_lock);
+		WRITE_ONCE(free_osk->sk_user_data, NULL);
+		write_unlock_bh(&free_osk->sk_callback_lock);
+	}
+
+	spin_unlock_bh(&reuseport_lock);
+put_file:
+	fput(socket->file);
+	return err;
+}
+
+/* Called from syscall */
+static int reuseport_array_get_next_key(struct bpf_map *map, void *key,
+					void *next_key)
+{
+	struct reuseport_array *array = reuseport_array(map);
+	u32 index = key ? *(u32 *)key : U32_MAX;
+	u32 *next = (u32 *)next_key;
+
+	if (index >= array->map.max_entries) {
+		*next = 0;
+		return 0;
+	}
+
+	if (index == array->map.max_entries - 1)
+		return -ENOENT;
+
+	*next = index + 1;
+	return 0;
+}
+
+const struct bpf_map_ops reuseport_array_ops = {
+	.map_alloc_check = reuseport_array_alloc_check,
+	.map_alloc = reuseport_array_alloc,
+	.map_free = reuseport_array_free,
+	.map_lookup_elem = reuseport_array_lookup_elem,
+	.map_get_next_key = reuseport_array_get_next_key,
+	.map_delete_elem = reuseport_array_delete_elem,
+};

kernel/bpf/syscall.c

@@ -684,6 +684,8 @@ static int map_lookup_elem(union bpf_attr *attr)
 		err = bpf_fd_array_map_lookup_elem(map, key, value);
 	} else if (IS_FD_HASH(map)) {
 		err = bpf_fd_htab_map_lookup_elem(map, key, value);
+	} else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) {
+		err = bpf_fd_reuseport_array_lookup_elem(map, key, value);
 	} else {
 		rcu_read_lock();
 		ptr = map->ops->map_lookup_elem(map, key);
@@ -790,6 +792,10 @@ static int map_update_elem(union bpf_attr *attr)
 		err = bpf_fd_htab_map_update_elem(map, f.file, key, value,
 						  attr->flags);
 		rcu_read_unlock();
+	} else if (map->map_type == BPF_MAP_TYPE_REUSEPORT_SOCKARRAY) {
+		/* rcu_read_lock() is not needed */
+		err = bpf_fd_reuseport_array_update_elem(map, key, value,
+							 attr->flags);
 	} else {
 		rcu_read_lock();
 		err = map->ops->map_update_elem(map, key, value, attr->flags);

net/core/sock_reuseport.c

@@ -186,6 +186,14 @@ void reuseport_detach_sock(struct sock *sk)
 	spin_lock_bh(&reuseport_lock);
 	reuse = rcu_dereference_protected(sk->sk_reuseport_cb,
 					  lockdep_is_held(&reuseport_lock));
+
+	/* At least one of the sk in this reuseport group is added to
+	 * a bpf map.  Notify the bpf side.  The bpf map logic will
+	 * remove the sk if it is indeed added to a bpf map.
+	 */
+	if (reuse->reuseport_id)
+		bpf_sk_reuseport_detach(sk);
+
 	rcu_assign_pointer(sk->sk_reuseport_cb, NULL);
 
 	for (i = 0; i < reuse->num_socks; i++) {