bpf: map_get_next_key to return first key on NULL

When iterating through a map, we need to find a key that does not exist
in the map so map_get_next_key will give us the first key of the map.
This often requires a lot of guessing in production systems.

This patch makes map_get_next_key return the first key when the key
pointer in the parameter is NULL.
Signed-off-by: Teng Qin <qinteng@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/trace/events/bpf.h

@@ -321,11 +321,14 @@ TRACE_EVENT(bpf_map_next_key,
 		__dynamic_array(u8, key, map->key_size)
 		__dynamic_array(u8, nxt, map->key_size)
 		__field(bool, key_trunc)
+		__field(bool, key_null)
 		__field(int, ufd)
 	),
 
 	TP_fast_assign(
+		if (key)
 			memcpy(__get_dynamic_array(key), key, map->key_size);
+		__entry->key_null = !key;
 		memcpy(__get_dynamic_array(nxt), key_next, map->key_size);
 		__entry->type      = map->map_type;
 		__entry->key_len   = min(map->key_size, 16U);
@@ -336,8 +339,9 @@ TRACE_EVENT(bpf_map_next_key,
 	TP_printk("map type=%s ufd=%d key=[%s%s] next=[%s%s]",
 		  __print_symbolic(__entry->type, __MAP_TYPE_SYM_TAB),
 		  __entry->ufd,
-		  __print_hex(__get_dynamic_array(key), __entry->key_len),
-		  __entry->key_trunc ? " ..." : "",
+		  __entry->key_null ? "NULL" : __print_hex(__get_dynamic_array(key),
+							   __entry->key_len),
+		  __entry->key_trunc && !__entry->key_null ? " ..." : "",
 		  __print_hex(__get_dynamic_array(nxt), __entry->key_len),
 		  __entry->key_trunc ? " ..." : "")
 );

kernel/bpf/arraymap.c

@@ -182,7 +182,7 @@ int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value)
 static int array_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
 {
 	struct bpf_array *array = container_of(map, struct bpf_array, map);
-	u32 index = *(u32 *)key;
+	u32 index = key ? *(u32 *)key : U32_MAX;
 	u32 *next = (u32 *)next_key;
 
 	if (index >= array->map.max_entries) {

kernel/bpf/hashtab.c

@@ -540,12 +540,15 @@ static int htab_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
 	struct hlist_nulls_head *head;
 	struct htab_elem *l, *next_l;
 	u32 hash, key_size;
-	int i;
+	int i = 0;
 
 	WARN_ON_ONCE(!rcu_read_lock_held());
 
 	key_size = map->key_size;
 
+	if (!key)
+		goto find_first_elem;
+
 	hash = htab_map_hash(key, key_size);
 
 	head = select_bucket(htab, hash);
@@ -553,10 +556,8 @@ static int htab_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
 	/* lookup the key */
 	l = lookup_nulls_elem_raw(head, hash, key, key_size, htab->n_buckets);
 
-	if (!l) {
-		i = 0;
+	if (!l)
 		goto find_first_elem;
-	}
 
 	/* key was found, get next key in the same bucket */
 	next_l = hlist_nulls_entry_safe(rcu_dereference_raw(hlist_nulls_next_rcu(&l->hash_node)),

kernel/bpf/syscall.c

@@ -536,6 +536,7 @@ static int map_get_next_key(union bpf_attr *attr)
 	if (IS_ERR(map))
 		return PTR_ERR(map);
 
+	if (ukey) {
 		err = -ENOMEM;
 		key = kmalloc(map->key_size, GFP_USER);
 		if (!key)
@@ -544,6 +545,9 @@ static int map_get_next_key(union bpf_attr *attr)
 		err = -EFAULT;
 		if (copy_from_user(key, ukey, map->key_size) != 0)
 			goto free_key;
+	} else {
+		key = NULL;
+	}
 
 	err = -ENOMEM;
 	next_key = kmalloc(map->key_size, GFP_USER);

tools/testing/selftests/bpf/test_maps.c

@@ -28,7 +28,7 @@ static int map_flags;
 
 static void test_hashmap(int task, void *data)
 {
-	long long key, next_key, value;
+	long long key, next_key, first_key, value;
 	int fd;
 
 	fd = bpf_create_map(BPF_MAP_TYPE_HASH, sizeof(key), sizeof(value),
@@ -89,10 +89,13 @@ static void test_hashmap(int task, void *data)
 	assert(bpf_map_delete_elem(fd, &key) == -1 && errno == ENOENT);
 
 	/* Iterate over two elements. */
+	assert(bpf_map_get_next_key(fd, NULL, &first_key) == 0 &&
+	       (first_key == 1 || first_key == 2));
 	assert(bpf_map_get_next_key(fd, &key, &next_key) == 0 &&
-	       (next_key == 1 || next_key == 2));
+	       (next_key == first_key));
 	assert(bpf_map_get_next_key(fd, &next_key, &next_key) == 0 &&
-	       (next_key == 1 || next_key == 2));
+	       (next_key == 1 || next_key == 2) &&
+	       (next_key != first_key));
 	assert(bpf_map_get_next_key(fd, &next_key, &next_key) == -1 &&
 	       errno == ENOENT);
 
@@ -105,6 +108,8 @@ static void test_hashmap(int task, void *data)
 
 	key = 0;
 	/* Check that map is empty. */
+	assert(bpf_map_get_next_key(fd, NULL, &next_key) == -1 &&
+	       errno == ENOENT);
 	assert(bpf_map_get_next_key(fd, &key, &next_key) == -1 &&
 	       errno == ENOENT);
 
@@ -133,7 +138,7 @@ static void test_hashmap_percpu(int task, void *data)
 {
 	unsigned int nr_cpus = bpf_num_possible_cpus();
 	long long value[nr_cpus];
-	long long key, next_key;
+	long long key, next_key, first_key;
 	int expected_key_mask = 0;
 	int fd, i;
 
@@ -193,7 +198,13 @@ static void test_hashmap_percpu(int task, void *data)
 	assert(bpf_map_delete_elem(fd, &key) == -1 && errno == ENOENT);
 
 	/* Iterate over two elements. */
+	assert(bpf_map_get_next_key(fd, NULL, &first_key) == 0 &&
+	       ((expected_key_mask & first_key) == first_key));
 	while (!bpf_map_get_next_key(fd, &key, &next_key)) {
+		if (first_key) {
+			assert(next_key == first_key);
+			first_key = 0;
+		}
 		assert((expected_key_mask & next_key) == next_key);
 		expected_key_mask &= ~next_key;
 
@@ -219,6 +230,8 @@ static void test_hashmap_percpu(int task, void *data)
 
 	key = 0;
 	/* Check that map is empty. */
+	assert(bpf_map_get_next_key(fd, NULL, &next_key) == -1 &&
+	       errno == ENOENT);
 	assert(bpf_map_get_next_key(fd, &key, &next_key) == -1 &&
 	       errno == ENOENT);
 
@@ -264,6 +277,8 @@ static void test_arraymap(int task, void *data)
 	assert(bpf_map_lookup_elem(fd, &key, &value) == -1 && errno == ENOENT);
 
 	/* Iterate over two elements. */
+	assert(bpf_map_get_next_key(fd, NULL, &next_key) == 0 &&
+	       next_key == 0);
 	assert(bpf_map_get_next_key(fd, &key, &next_key) == 0 &&
 	       next_key == 0);
 	assert(bpf_map_get_next_key(fd, &next_key, &next_key) == 0 &&
@@ -319,6 +334,8 @@ static void test_arraymap_percpu(int task, void *data)
 	assert(bpf_map_lookup_elem(fd, &key, values) == -1 && errno == ENOENT);
 
 	/* Iterate over two elements. */
+	assert(bpf_map_get_next_key(fd, NULL, &next_key) == 0 &&
+	       next_key == 0);
 	assert(bpf_map_get_next_key(fd, &key, &next_key) == 0 &&
 	       next_key == 0);
 	assert(bpf_map_get_next_key(fd, &next_key, &next_key) == 0 &&
@@ -400,6 +417,8 @@ static void test_map_large(void)
 	       errno == E2BIG);
 
 	/* Iterate through all elements. */
+	assert(bpf_map_get_next_key(fd, NULL, &key) == 0);
+	key.c = -1;
 	for (i = 0; i < MAP_SIZE; i++)
 		assert(bpf_map_get_next_key(fd, &key, &key) == 0);
 	assert(bpf_map_get_next_key(fd, &key, &key) == -1 && errno == ENOENT);
@@ -499,6 +518,7 @@ static void test_map_parallel(void)
 	       errno == EEXIST);
 
 	/* Check that all elements were inserted. */
+	assert(bpf_map_get_next_key(fd, NULL, &key) == 0);
 	key = -1;
 	for (i = 0; i < MAP_SIZE; i++)
 		assert(bpf_map_get_next_key(fd, &key, &key) == 0);
@@ -518,6 +538,7 @@ static void test_map_parallel(void)
 
 	/* Nothing should be left. */
 	key = -1;
+	assert(bpf_map_get_next_key(fd, NULL, &key) == -1 && errno == ENOENT);
 	assert(bpf_map_get_next_key(fd, &key, &key) == -1 && errno == ENOENT);
 }