libbpf: Add capability for resizing datasec maps

This patch updates bpf_map__set_value_size() so that if the given map is
memory mapped, it will attempt to resize the mapped region. Initial
contents of the mapped region are preserved. BTF is not required, but
after the mapping is resized an attempt is made to adjust the associated
BTF information if the following criteria is met:
 - BTF info is present
 - the map is a datasec
 - the final variable in the datasec is an array

... the resulting BTF info will be updated so that the final array
variable is associated with a new BTF array type sized to cover the
requested size.

Note that the initial resizing of the memory mapped region can succeed
while the subsequent BTF adjustment can fail. In this case, BTF info is
dropped from the map by clearing the key and value type.

Signed-off-by: JP Kobryn <inwardvessel@gmail.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/bpf/20230524004537.18614-2-inwardvessel@gmail.com

2 files changed, 142 insertions, 11 deletions

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index ad1ec893b41b..5cca00979aae 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -1500,16 +1500,36 @@ static struct bpf_map *bpf_object__add_map(struct bpf_object *obj)
 	return map;
 }
 
-static size_t bpf_map_mmap_sz(const struct bpf_map *map)
+static size_t bpf_map_mmap_sz(unsigned int value_sz, unsigned int max_entries)
 {
-	long page_sz = sysconf(_SC_PAGE_SIZE);
+	const long page_sz = sysconf(_SC_PAGE_SIZE);
 	size_t map_sz;
 
-	map_sz = (size_t)roundup(map->def.value_size, 8) * map->def.max_entries;
+	map_sz = (size_t)roundup(value_sz, 8) * max_entries;
 	map_sz = roundup(map_sz, page_sz);
 	return map_sz;
 }
 
+static int bpf_map_mmap_resize(struct bpf_map *map, size_t old_sz, size_t new_sz)
+{
+	void *mmaped;
+
+	if (!map->mmaped)
+		return -EINVAL;
+
+	if (old_sz == new_sz)
+		return 0;
+
+	mmaped = mmap(NULL, new_sz, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_ANONYMOUS, -1, 0);
+	if (mmaped == MAP_FAILED)
+		return -errno;
+
+	memcpy(mmaped, map->mmaped, min(old_sz, new_sz));
+	munmap(map->mmaped, old_sz);
+	map->mmaped = mmaped;
+	return 0;
+}
+
 static char *internal_map_name(struct bpf_object *obj, const char *real_name)
 {
 	char map_name[BPF_OBJ_NAME_LEN], *p;
@@ -1608,6 +1628,7 @@ bpf_object__init_internal_map(struct bpf_object *obj, enum libbpf_map_type type,
 {
 	struct bpf_map_def *def;
 	struct bpf_map *map;
+	size_t mmap_sz;
 	int err;
 
 	map = bpf_object__add_map(obj);
@@ -1642,7 +1663,8 @@ bpf_object__init_internal_map(struct bpf_object *obj, enum libbpf_map_type type,
 	pr_debug("map '%s' (global data): at sec_idx %d, offset %zu, flags %x.\n",
 		 map->name, map->sec_idx, map->sec_offset, def->map_flags);
 
-	map->mmaped = mmap(NULL, bpf_map_mmap_sz(map), PROT_READ | PROT_WRITE,
+	mmap_sz = bpf_map_mmap_sz(map->def.value_size, map->def.max_entries);
+	map->mmaped = mmap(NULL, mmap_sz, PROT_READ | PROT_WRITE,
 			   MAP_SHARED | MAP_ANONYMOUS, -1, 0);
 	if (map->mmaped == MAP_FAILED) {
 		err = -errno;
@@ -8294,7 +8316,10 @@ static void bpf_map__destroy(struct bpf_map *map)
 	map->init_slots_sz = 0;
 
 	if (map->mmaped) {
-		munmap(map->mmaped, bpf_map_mmap_sz(map));
+		size_t mmap_sz;
+
+		mmap_sz = bpf_map_mmap_sz(map->def.value_size, map->def.max_entries);
+		munmap(map->mmaped, mmap_sz);
 		map->mmaped = NULL;
 	}
 
@@ -9412,10 +9437,103 @@ __u32 bpf_map__value_size(const struct bpf_map *map)
 	return map->def.value_size;
 }
 
+static int map_btf_datasec_resize(struct bpf_map *map, __u32 size)
+{
+	struct btf *btf;
+	struct btf_type *datasec_type, *var_type;
+	struct btf_var_secinfo *var;
+	const struct btf_type *array_type;
+	const struct btf_array *array;
+	int vlen, element_sz;
+	__u32 nr_elements, new_array_id;
+
+	/* check btf existence */
+	btf = bpf_object__btf(map->obj);
+	if (!btf)
+		return -ENOENT;
+
+	/* verify map is datasec */
+	datasec_type = btf_type_by_id(btf, bpf_map__btf_value_type_id(map));
+	if (!btf_is_datasec(datasec_type)) {
+		pr_warn("map '%s': cannot be resized, map value type is not a datasec\n",
+			bpf_map__name(map));
+		return -EINVAL;
+	}
+
+	/* verify datasec has at least one var */
+	vlen = btf_vlen(datasec_type);
+	if (vlen == 0) {
+		pr_warn("map '%s': cannot be resized, map value datasec is empty\n",
+			bpf_map__name(map));
+		return -EINVAL;
+	}
+
+	/* verify last var in the datasec is an array */
+	var = &btf_var_secinfos(datasec_type)[vlen - 1];
+	var_type = btf_type_by_id(btf, var->type);
+	array_type = skip_mods_and_typedefs(btf, var_type->type, NULL);
+	if (!btf_is_array(array_type)) {
+		pr_warn("map '%s': cannot be resized, last var must be an array\n",
+			bpf_map__name(map));
+		return -EINVAL;
+	}
+
+	/* verify request size aligns with array */
+	array = btf_array(array_type);
+	element_sz = btf__resolve_size(btf, array->type);
+	if (element_sz <= 0 || (size - var->offset) % element_sz != 0) {
+		pr_warn("map '%s': cannot be resized, element size (%d) doesn't align with new total size (%u)\n",
+			bpf_map__name(map), element_sz, size);
+		return -EINVAL;
+	}
+
+	/* create a new array based on the existing array, but with new length */
+	nr_elements = (size - var->offset) / element_sz;
+	new_array_id = btf__add_array(btf, array->index_type, array->type, nr_elements);
+	if (new_array_id < 0)
+		return new_array_id;
+
+	/* adding a new btf type invalidates existing pointers to btf objects,
+	 * so refresh pointers before proceeding
+	 */
+	datasec_type = btf_type_by_id(btf, map->btf_value_type_id);
+	var = &btf_var_secinfos(datasec_type)[vlen - 1];
+	var_type = btf_type_by_id(btf, var->type);
+
+	/* finally update btf info */
+	datasec_type->size = size;
+	var->size = size - var->offset;
+	var_type->type = new_array_id;
+
+	return 0;
+}
+
 int bpf_map__set_value_size(struct bpf_map *map, __u32 size)
 {
 	if (map->fd >= 0)
 		return libbpf_err(-EBUSY);
+
+	if (map->mmaped) {
+		int err;
+		size_t mmap_old_sz, mmap_new_sz;
+
+		mmap_old_sz = bpf_map_mmap_sz(map->def.value_size, map->def.max_entries);
+		mmap_new_sz = bpf_map_mmap_sz(size, map->def.max_entries);
+		err = bpf_map_mmap_resize(map, mmap_old_sz, mmap_new_sz);
+		if (err) {
+			pr_warn("map '%s': failed to resize memory-mapped region: %d\n",
+				bpf_map__name(map), err);
+			return err;
+		}
+		err = map_btf_datasec_resize(map, size);
+		if (err && err != -ENOENT) {
+			pr_warn("map '%s': failed to adjust resized BTF, clearing BTF key/value info: %d\n",
+				bpf_map__name(map), err);
+			map->btf_value_type_id = 0;
+			map->btf_key_type_id = 0;
+		}
+	}
+
 	map->def.value_size = size;
 	return 0;
 }
@@ -9441,7 +9559,7 @@ int bpf_map__set_initial_value(struct bpf_map *map,
 	return 0;
 }
 
-const void *bpf_map__initial_value(struct bpf_map *map, size_t *psize)
+void *bpf_map__initial_value(struct bpf_map *map, size_t *psize)
 {
 	if (!map->mmaped)
 		return NULL;
@@ -12693,7 +12811,7 @@ int bpf_object__load_skeleton(struct bpf_object_skeleton *s)
 
 	for (i = 0; i < s->map_cnt; i++) {
 		struct bpf_map *map = *s->maps[i].map;
-		size_t mmap_sz = bpf_map_mmap_sz(map);
+		size_t mmap_sz = bpf_map_mmap_sz(map->def.value_size, map->def.max_entries);
 		int prot, map_fd = bpf_map__fd(map);
 		void **mmaped = s->maps[i].mmaped;
 
@@ -12720,8 +12838,7 @@ int bpf_object__load_skeleton(struct bpf_object_skeleton *s)
 		 * as per normal clean up procedure, so we don't need to worry
 		 * about it from skeleton's clean up perspective.
 		 */
-		*mmaped = mmap(map->mmaped, mmap_sz, prot,
-				MAP_SHARED | MAP_FIXED, map_fd, 0);
+		*mmaped = mmap(map->mmaped, mmap_sz, prot, MAP_SHARED | MAP_FIXED, map_fd, 0);
 		if (*mmaped == MAP_FAILED) {
 			err = -errno;
 			*mmaped = NULL;
diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
index 0b7362397ea3..754da73c643b 100644
--- a/tools/lib/bpf/libbpf.h
+++ b/tools/lib/bpf/libbpf.h
@@ -869,8 +869,22 @@ LIBBPF_API int bpf_map__set_numa_node(struct bpf_map *map, __u32 numa_node);
 /* get/set map key size */
 LIBBPF_API __u32 bpf_map__key_size(const struct bpf_map *map);
 LIBBPF_API int bpf_map__set_key_size(struct bpf_map *map, __u32 size);
-/* get/set map value size */
+/* get map value size */
 LIBBPF_API __u32 bpf_map__value_size(const struct bpf_map *map);
+/**
+ * @brief **bpf_map__set_value_size()** sets map value size.
+ * @param map the BPF map instance
+ * @return 0, on success; negative error, otherwise
+ *
+ * There is a special case for maps with associated memory-mapped regions, like
+ * the global data section maps (bss, data, rodata). When this function is used
+ * on such a map, the mapped region is resized. Afterward, an attempt is made to
+ * adjust the corresponding BTF info. This attempt is best-effort and can only
+ * succeed if the last variable of the data section map is an array. The array
+ * BTF type is replaced by a new BTF array type with a different length.
+ * Any previously existing pointers returned from bpf_map__initial_value() or
+ * corresponding data section skeleton pointer must be reinitialized.
+ */
 LIBBPF_API int bpf_map__set_value_size(struct bpf_map *map, __u32 size);
 /* get map key/value BTF type IDs */
 LIBBPF_API __u32 bpf_map__btf_key_type_id(const struct bpf_map *map);
@@ -884,7 +898,7 @@ LIBBPF_API int bpf_map__set_map_extra(struct bpf_map *map, __u64 map_extra);
 
 LIBBPF_API int bpf_map__set_initial_value(struct bpf_map *map,
 					  const void *data, size_t size);
-LIBBPF_API const void *bpf_map__initial_value(struct bpf_map *map, size_t *psize);
+LIBBPF_API void *bpf_map__initial_value(struct bpf_map *map, size_t *psize);
 
 /**
  * @brief **bpf_map__is_internal()** tells the caller whether or not the