Re: [PATCH v7 5/6] kernfs: Use RCU to access kernfs_node::parent.

[Yonghong Song]

On 2/10/25 12:43 AM, Sebastian Andrzej Siewior wrote:
> On 2025-02-03 14:50:22 [+0100], To cgroups@xxxxxxxxxxxxxxx wrote:
> > kernfs_rename_lock is used to obtain stable kernfs_node::{name|parent}
> > pointer. This is a preparation to access kernfs_node::parent under RCU
> > and ensure that the pointer remains stable under the RCU lifetime
> > guarantees.
> …
>
> The robot complained that the selftests for bpf broke. As it turns out,
> the tests access kernfs_node::{parent|name} and after the rename
> ::parent is gone so it does not compile.
> If there are no objections, I would merge this into 5/6 and repost.
> "test_progs -a test_profiler" passes.
>
> diff --git a/tools/testing/selftests/bpf/progs/profiler.inc.h b/tools/testing/selftests/bpf/progs/profiler.inc.h
> index 8bd1ebd7d6afd..a4f518ee5f4de 100644
> --- a/tools/testing/selftests/bpf/progs/profiler.inc.h
> +++ b/tools/testing/selftests/bpf/progs/profiler.inc.h
> @@ -223,7 +223,7 @@ static INLINE void* read_full_cgroup_path(struct kernfs_node* cgroup_node,
>   		if (bpf_cmp_likely(filepart_length, <=, MAX_PATH)) {
>   			payload += filepart_length;
>   		}
> -		cgroup_node = BPF_CORE_READ(cgroup_node, parent);
> +		cgroup_node = BPF_CORE_READ(cgroup_node, __parent);
>   	}
>   	return payload;
>   }
> @@ -300,6 +300,7 @@ static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
>   	cgroup_data->cgroup_proc_length = 0;
>   	cgroup_data->cgroup_full_length = 0;
>   
> +	bpf_rcu_read_lock();
>   	size_t cgroup_root_length =
>   		bpf_probe_read_kernel_str(payload, MAX_PATH,
>   					  BPF_CORE_READ(root_kernfs, name));
> @@ -323,6 +324,7 @@ static INLINE void* populate_cgroup_info(struct cgroup_data_t* cgroup_data,
>   		cgroup_data->cgroup_full_length = payload_end_pos - payload;
>   		payload = payload_end_pos;
>   	}
> +	bpf_rcu_read_unlock();

All programs calling this function populate_cgroup_info() is not sleepable program
so the whole prog is protected by rcu and there is no need for above
bpf_rcu_read_{lock,unlock}().

>   
>   	return (void*)payload;
>   }