Re: [RFC PATCH bpf-next 3/3] selftests/bpf: Add selftests for cgroup1 local storage

[Yonghong Song <yonghong.song@linux.dev>]

On 12/5/23 9:37 AM, Yafang Shao wrote:
> Expanding the test coverage from cgroup2 to include cgroup1. The result
> as follows,
>
> Already existing test cases for cgroup2:
>    #48/1    cgrp_local_storage/tp_btf:OK
>    #48/2    cgrp_local_storage/attach_cgroup:OK
>    #48/3    cgrp_local_storage/recursion:OK
>    #48/4    cgrp_local_storage/negative:OK
>    #48/5    cgrp_local_storage/cgroup_iter_sleepable:OK
>    #48/6    cgrp_local_storage/yes_rcu_lock:OK
>    #48/7    cgrp_local_storage/no_rcu_lock:OK
>
> Expanded test cases for cgroup1:
>    #48/8    cgrp_local_storage/cgrp1_tp_btf:OK
>    #48/9    cgrp_local_storage/cgrp1_recursion:OK
>    #48/10   cgrp_local_storage/cgrp1_negative:OK
>    #48/11   cgrp_local_storage/cgrp1_iter_sleepable:OK
>    #48/12   cgrp_local_storage/cgrp1_yes_rcu_lock:OK
>    #48/13   cgrp_local_storage/cgrp1_no_rcu_lock:OK
>
> Summary:
>    #48      cgrp_local_storage:OK
>    Summary: 1/13 PASSED, 0 SKIPPED, 0 FAILED
>
> Signed-off-by: Yafang Shao <laoar.shao@gmail.com>

LGTM with a few nits below.

Acked-by: Yonghong Song <yonghong.song@linux.dev>

> ---
>   .../bpf/prog_tests/cgrp_local_storage.c       | 92 ++++++++++++++++++-
>   .../selftests/bpf/progs/cgrp_ls_recursion.c   | 84 +++++++++++++----
>   .../selftests/bpf/progs/cgrp_ls_sleepable.c   | 67 ++++++++++++--
>   .../selftests/bpf/progs/cgrp_ls_tp_btf.c      | 82 ++++++++++++-----
>   4 files changed, 278 insertions(+), 47 deletions(-)
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c b/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c
> index 63e776f4176e..9524cde4cbf6 100644
> --- a/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c
> +++ b/tools/testing/selftests/bpf/prog_tests/cgrp_local_storage.c
> @@ -19,6 +19,9 @@ struct socket_cookie {
>   	__u64 cookie_value;
>   };
>   
> +bool is_cgroup1;
> +int target_hid;
> +
>   static void test_tp_btf(int cgroup_fd)
>   {
>   	struct cgrp_ls_tp_btf *skel;
> @@ -29,6 +32,9 @@ static void test_tp_btf(int cgroup_fd)
>   	if (!ASSERT_OK_PTR(skel, "skel_open_and_load"))
>   		return;
>   
> +	skel->bss->is_cgroup1 = is_cgroup1;
> +	skel->bss->target_hid = target_hid;

Let reverse the order like below to be consistent with other code patterns:
+	skel->bss->target_hid = target_hid;
+	skel->bss->is_cgroup1 = is_cgroup1;

> +
>   	/* populate a value in map_b */
>   	err = bpf_map_update_elem(bpf_map__fd(skel->maps.map_b), &cgroup_fd, &val1, BPF_ANY);
>   	if (!ASSERT_OK(err, "map_update_elem"))
> @@ -130,6 +136,9 @@ static void test_recursion(int cgroup_fd)
>   	if (!ASSERT_OK_PTR(skel, "skel_open_and_load"))
>   		return;
>   
> +	skel->bss->target_hid = target_hid;
> +	skel->bss->is_cgroup1 = is_cgroup1;
> +
>   	err = cgrp_ls_recursion__attach(skel);
>   	if (!ASSERT_OK(err, "skel_attach"))
>   		goto out;
> [...]
> diff --git a/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c b/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c
> index 4c7844e1dbfa..985ff419249c 100644
> --- a/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c
> +++ b/tools/testing/selftests/bpf/progs/cgrp_ls_sleepable.c
> @@ -17,7 +17,11 @@ struct {
>   
>   __u32 target_pid;
>   __u64 cgroup_id;
> +int target_hid;
> +bool is_cgroup1;
>   
> +struct cgroup *bpf_task_get_cgroup1(struct task_struct *task, int hierarchy_id) __ksym;
> +void bpf_cgroup_release(struct cgroup *cgrp) __ksym;
>   void bpf_rcu_read_lock(void) __ksym;
>   void bpf_rcu_read_unlock(void) __ksym;
>   
> @@ -37,23 +41,56 @@ int cgroup_iter(struct bpf_iter__cgroup *ctx)
>   	return 0;
>   }
>   
> +static void __no_rcu_lock(struct cgroup *cgrp)
> +{
> +	long *ptr;
> +
> +	/* Note that trace rcu is held in sleepable prog, so we can use
> +	 * bpf_cgrp_storage_get() in sleepable prog.
> +	 */
> +	ptr = bpf_cgrp_storage_get(&map_a, cgrp, 0,
> +				   BPF_LOCAL_STORAGE_GET_F_CREATE);
> +	if (ptr)
> +		cgroup_id = cgrp->kn->id;
> +}
> +
>   SEC("?fentry.s/" SYS_PREFIX "sys_getpgid")
> -int no_rcu_lock(void *ctx)
> +int cgrp1_no_rcu_lock(void *ctx)
>   {
>   	struct task_struct *task;
>   	struct cgroup *cgrp;
> -	long *ptr;
> +
> +	if (!is_cgroup1)
> +		return 0;

Do we need this check? Looks like the user space controls whether it will
be loaded or not depending on whether it is cgrp1.

> +
> +	task = bpf_get_current_task_btf();
> +	if (task->pid != target_pid)
> +		return 0;
> +
> +	/* bpf_task_get_cgroup1 can work in sleepable prog */
> +	cgrp = bpf_task_get_cgroup1(task, target_hid);
> +	if (!cgrp)
> +		return 0;
> +
> +	__no_rcu_lock(cgrp);
> +	bpf_cgroup_release(cgrp);
> +	return 0;
> +}
> +
> +SEC("?fentry.s/" SYS_PREFIX "sys_getpgid")
> +int no_rcu_lock(void *ctx)
> +{
> +	struct task_struct *task;
> +
> +	if (is_cgroup1)
> +		return 0;

Same here, check is not needed.

>   
>   	task = bpf_get_current_task_btf();
>   	if (task->pid != target_pid)
>   		return 0;
>   
>   	/* task->cgroups is untrusted in sleepable prog outside of RCU CS */
> -	cgrp = task->cgroups->dfl_cgrp;
> -	ptr = bpf_cgrp_storage_get(&map_a, cgrp, 0,
> -				   BPF_LOCAL_STORAGE_GET_F_CREATE);
> -	if (ptr)
> -		cgroup_id = cgrp->kn->id;
> +	__no_rcu_lock(task->cgroups->dfl_cgrp);
>   	return 0;
>   }
>   
> @@ -68,6 +105,22 @@ int yes_rcu_lock(void *ctx)
>   	if (task->pid != target_pid)
>   		return 0;
>   
> +	if (is_cgroup1) {
> +		bpf_rcu_read_lock();
> +		cgrp = bpf_task_get_cgroup1(task, target_hid);
> +		if (!cgrp) {
> +			bpf_rcu_read_unlock();
> +			return 0;
> +		}
> +
> +		ptr = bpf_cgrp_storage_get(&map_a, cgrp, 0, BPF_LOCAL_STORAGE_GET_F_CREATE);
> +		if (ptr)
> +			cgroup_id = cgrp->kn->id;
> +		bpf_cgroup_release(cgrp);
> +		bpf_rcu_read_unlock();
> +		return 0;
> +	}
> +
>   	bpf_rcu_read_lock();
>   	cgrp = task->cgroups->dfl_cgrp;
>   	/* cgrp is trusted under RCU CS */

[...]