On 10/5/20 6:41 AM, Luigi Rizzo wrote:
bpf_program__set_attach_target() will always fail with fd=0 (attach to a
kernel symbol) because obj->btf_vmlinux is NULL and there is no way to
set it.
Fix this by explicitly calling libbpf_find_kernel_btf() in the function.
Signed-off-by: Luigi Rizzo <lrizzo@xxxxxxxxxx>
---
tools/lib/bpf/libbpf.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index a4f55f8a460d..3a63db86666f 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -10352,10 +10352,13 @@ int bpf_program__set_attach_target(struct bpf_program *prog,
if (attach_prog_fd)
btf_id = libbpf_find_prog_btf_id(attach_func_name,
attach_prog_fd);
- else
- btf_id = __find_vmlinux_btf_id(prog->obj->btf_vmlinux,
- attach_func_name,
+ else {
+ struct btf *btf = libbpf_find_kernel_btf();
+
+ btf_id = __find_vmlinux_btf_id(btf, attach_func_name,
prog->expected_attach_type);
+ btf__free(btf);
+ }
Doesn't feel this is right fix. In libbpf, we have API function
libbpf_find_vmlinux_btf_id
just doing the above.
Also, could you point out why btf_vmlinux is NULL? Is this related
to fd = 0? We probably
should fix the problem there since what if other API functions
like bpf_program__set_attach_target() wants btf_vmlinux. We should
not duplicate this logic to other API functions.
if (btf_id < 0)
return btf_id;
