On 7/2/20 10:25 AM, John Fastabend wrote:
Fix segfault from bpftool by adding emit_obj_refs_plain when skeleton
code is disabled.
Tested by deleting BUILD_BPF_SKELS in Makefile.
You probably hit this issue with a different way, right?
Old clang, or anything? I would be good to clarify in
the commit message. People in general do not tweak Makefile
to find bugs, I guess.
# ./bpftool prog show
Error: bpftool built without PID iterator support
3: cgroup_skb tag 7be49e3934a125ba gpl
loaded_at 2020-07-01T08:01:29-0700 uid 0
Segmentation fault
Signed-off-by: John Fastabend <john.fastabend@xxxxxxxxx>
Ack with request for better description in the commit message.
Acked-by: Yonghong Song <yhs@xxxxxx>
---
tools/bpf/bpftool/pids.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/tools/bpf/bpftool/pids.c b/tools/bpf/bpftool/pids.c
index 2709be4de2b1..7d5416667c85 100644
--- a/tools/bpf/bpftool/pids.c
+++ b/tools/bpf/bpftool/pids.c
@@ -19,6 +19,7 @@ int build_obj_refs_table(struct obj_refs_table *table, enum bpf_obj_type type)
return -ENOTSUP;
}
void delete_obj_refs_table(struct obj_refs_table *table) {}
+void emit_obj_refs_plain(struct obj_refs_table *table, __u32 id, const char *prefix) {}
#else /* BPFTOOL_WITHOUT_SKELETONS */