When you collect a callgraph profile with operf, the opreport output incorrectly implies recursive calls. For example, my simple memcpy testcase does the following:
main -> do_my_memcpy -> memcpy (libc)
The part of the callgraph report that focuses on do_my_memcpy looks like the following:
4757 50.0000 memcpyt do_my_memcpy 4757 50.0000 memcpyt main 4757 6.3185 memcpyt do_my_memcpy 4757 49.9842 memcpyt do_my_memcpy 4757 49.9842 memcpyt do_my_memcpy [self] 3 0.0315 no-vmlinux /no-vmlinux
So it appears that do_my_memcpy calls itself, which it does not do.
If I use 'perf record' to get a callgraph profile, the 'perf report' looks like the following:
6.88% memcpyt memcpyt [.] do_my_memcpy | --- do_my_memcpy main __libc_start_main
So here, too, it seems to me that do_my_memcpy calls do_my_memcpy. But when I reported this issue to perf/perf_events kernel developers, they don't think it's a problem. I didn't push it because we can easily handle this in operf by dropping the first address in the callchain (since it's repeated in the second callchain entry) that the kernel returns to us.
Log in to post a comment.