[PATCH bpf] libbpf: Check if map names are not NULL

[PATCH bpf] libbpf: Check if map names are not NULL

[Michal Rostecki]

BPF maps do not have to be named, so map names can be NULL. Map name
pointers should be checked before doing any operations which cannot be
done on NULL pointers.

This issue was detected by the following errors coming from bpftool
built with AddressSanitizer:

bpf.c:256:2: runtime error: null pointer passed as argument 2, which is declared to never be null
bpf.c:92:2: runtime error: null pointer passed as argument 2, which is declared to never be null
bpf.c:169:2: runtime error: null pointer passed as argument 2, which is declared to never be null

Fixes: 88cda1c9da02 ("bpf: libbpf: Provide basic API support to specify BPF obj name")
Fixes: d7be143b67c2 ("libbpf: Support expected_attach_type at prog load")
Fixes: 8a138aed4a80 ("bpf: btf: Add BTF support to libbpf")
Signed-off-by: Michal Rostecki <mrostecki@opensuse.org>
---
 tools/lib/bpf/bpf.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/tools/lib/bpf/bpf.c b/tools/lib/bpf/bpf.c
index 9cd015574e83..21484ff1cecb 100644
--- a/tools/lib/bpf/bpf.c
+++ b/tools/lib/bpf/bpf.c
@@ -89,8 +89,9 @@ int bpf_create_map_xattr(const struct bpf_create_map_attr *create_attr)
 	attr.value_size = create_attr->value_size;
 	attr.max_entries = create_attr->max_entries;
 	attr.map_flags = create_attr->map_flags;
-	memcpy(attr.map_name, create_attr->name,
-	       min(name_len, BPF_OBJ_NAME_LEN - 1));
+	if (create_attr->name)
+		memcpy(attr.map_name, create_attr->name,
+		       min(name_len, BPF_OBJ_NAME_LEN - 1));
 	attr.numa_node = create_attr->numa_node;
 	attr.btf_fd = create_attr->btf_fd;
 	attr.btf_key_type_id = create_attr->btf_key_type_id;
@@ -166,7 +167,9 @@ int bpf_create_map_in_map_node(enum bpf_map_type map_type, const char *name,
 	attr.inner_map_fd = inner_map_fd;
 	attr.max_entries = max_entries;
 	attr.map_flags = map_flags;
-	memcpy(attr.map_name, name, min(name_len, BPF_OBJ_NAME_LEN - 1));
+	if (name)
+		memcpy(attr.map_name, name,
+		       min(name_len, BPF_OBJ_NAME_LEN - 1));
 
 	if (node >= 0) {
 		attr.map_flags |= BPF_F_NUMA_NODE;
@@ -253,8 +256,9 @@ int bpf_load_program_xattr(const struct bpf_load_program_attr *load_attr,
 	attr.line_info_rec_size = load_attr->line_info_rec_size;
 	attr.line_info_cnt = load_attr->line_info_cnt;
 	attr.line_info = ptr_to_u64(load_attr->line_info);
-	memcpy(attr.prog_name, load_attr->name,
-	       min(name_len, BPF_OBJ_NAME_LEN - 1));
+	if (load_attr->name)
+		memcpy(attr.prog_name, load_attr->name,
+		       min(name_len, BPF_OBJ_NAME_LEN - 1));
 
 	fd = sys_bpf_prog_load(&attr, sizeof(attr));
 	if (fd >= 0)
-- 
2.16.4

Re: [PATCH bpf] libbpf: Check if map names are not NULL

[Alexei Starovoitov]

On Wed, Mar 20, 2019 at 01:27:42PM +0100, Michal Rostecki wrote:
> BPF maps do not have to be named, so map names can be NULL. Map name
> pointers should be checked before doing any operations which cannot be
> done on NULL pointers.
> 
> This issue was detected by the following errors coming from bpftool
> built with AddressSanitizer:
> 
> bpf.c:256:2: runtime error: null pointer passed as argument 2, which is declared to never be null
> bpf.c:92:2: runtime error: null pointer passed as argument 2, which is declared to never be null
> bpf.c:169:2: runtime error: null pointer passed as argument 2, which is declared to never be null
> 
> Fixes: 88cda1c9da02 ("bpf: libbpf: Provide basic API support to specify BPF obj name")
> Fixes: d7be143b67c2 ("libbpf: Support expected_attach_type at prog load")
> Fixes: 8a138aed4a80 ("bpf: btf: Add BTF support to libbpf")
> Signed-off-by: Michal Rostecki <mrostecki@opensuse.org>

I don't think that's necessary. memcpy is fine with src==null and len==0