Re: [PATCH] bpf: Clean up all resources when register_fprobe_ips returns an error

[Jiri Olsa]

On Tue, Oct 25, 2022 at 04:06:28PM +0800, Chuang Wang wrote:
> When register_fprobe_ips returns an error, bpf_link_cleanup just cleans
> up bpf_link_primer's resources and forgets to clean up
> bpf_kprobe_multi_link, addrs, cookies.
> 
> So, by adding 'goto error', this ensures that all resources are cleaned
> up.
> 
> Cc: stable@xxxxxxxxxxxxxxx
> Fixes: 0dcac2725406 ("bpf: Add multi kprobe link")
> Signed-off-by: Chuang Wang <nashuiliang@xxxxxxxxx>
> ---
>  kernel/trace/bpf_trace.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> index 1ed08967fb97..5b806ef20857 100644
> --- a/kernel/trace/bpf_trace.c
> +++ b/kernel/trace/bpf_trace.c
> @@ -2778,7 +2778,7 @@ int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>  	err = register_fprobe_ips(&link->fp, addrs, cnt);
>  	if (err) {
>  		bpf_link_cleanup(&link_primer);
> -		return err;
> +		goto error;

that should be taken care of bpf_kprobe_multi_link_dealloc,
through the fput path in bpf_link_cleanup

do you see any related memory leak like report in kmemleak?

jirka

>  	}
>  
>  	return bpf_link_settle(&link_primer);
> -- 
> 2.34.1
>