On 20/10/26 03:25, Alexei Starovoitov wrote: > On Mon, Oct 19, 2026 at 08:36 AM Leon Hwang <[email protected]> wrote: >> + ret = bpf_log_attr_report_size(&attr_log, 0); >> + if (ret) >> + goto put_prog; > > can we drop bpf_log_attr_report_size() ? > Nothing is logged when the attach succeeds, so there is no size to > report, and libbpf zeroes attr_common, so log_true_size reads 0 anyway. > Finalizing only when ret < 0 already keeps -EFAULT away from a live link. > Then no need for the 2nd write, the log.c refactoring and the put_prog > label.
Sounds reasonable to me. Besides libbpf, other BPF loaders, e.g. cilium/ebpf, also need to zero attr_common before calling the BPF syscall. > > Something like this (compile tested only): Will change like this. Thanks, Leon > [...]
