OSDN Git Service

bpftool: Fix a leak of btf object
authorMartin KaFai Lau <kafai@fb.com>
Wed, 15 Jan 2020 23:00:19 +0000 (15:00 -0800)
committerAlexei Starovoitov <ast@kernel.org>
Wed, 15 Jan 2020 23:23:27 +0000 (15:23 -0800)
When testing a map has btf or not, maps_have_btf() tests it by actually
getting a btf_fd from sys_bpf(BPF_BTF_GET_FD_BY_ID). However, it
forgot to btf__free() it.

In maps_have_btf() stage, there is no need to test it by really
calling sys_bpf(BPF_BTF_GET_FD_BY_ID). Testing non zero
info.btf_id is good enough.

Also, the err_close case is unnecessary, and also causes double
close() because the calling func do_dump() will close() all fds again.

Fixes: 99f9863a0c45 ("bpftool: Match maps by name")
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Cc: Paul Chaignon <paul.chaignon@orange.com>
Link: https://lore.kernel.org/bpf/20200115230019.1101352-1-kafai@fb.com
tools/bpf/bpftool/map.c

index c01f76f..e00e9e1 100644 (file)
@@ -915,32 +915,20 @@ static int maps_have_btf(int *fds, int nb_fds)
 {
        struct bpf_map_info info = {};
        __u32 len = sizeof(info);
-       struct btf *btf = NULL;
        int err, i;
 
        for (i = 0; i < nb_fds; i++) {
                err = bpf_obj_get_info_by_fd(fds[i], &info, &len);
                if (err) {
                        p_err("can't get map info: %s", strerror(errno));
-                       goto err_close;
-               }
-
-               err = btf__get_from_id(info.btf_id, &btf);
-               if (err) {
-                       p_err("failed to get btf");
-                       goto err_close;
+                       return -1;
                }
 
-               if (!btf)
+               if (!info.btf_id)
                        return 0;
        }
 
        return 1;
-
-err_close:
-       for (; i < nb_fds; i++)
-               close(fds[i]);
-       return -1;
 }
 
 static int