On Sat, 15 Oct 2022 at 01:45, David Vernet <void@xxxxxxxxxxxxx> wrote:
>
> A previous change added a series of kfuncs for storing struct
> task_struct objects as referenced kptrs. This patch adds a new
> task_kfunc test suite for validating their expected behavior.
>
> Signed-off-by: David Vernet <void@xxxxxxxxxxxxx>
> ---
> [...]
> +
> +SEC("tp_btf/task_newtask")
> +int BPF_PROG(task_kfunc_acquire_trusted_nested, struct task_struct *task, u64 clone_flags)
> +{
> + struct task_struct *acquired;
> +
> + if (!is_test_kfunc_task())
> + return 0;
> +
> + /* Can't invoke bpf_task_acquire() on a trusted pointer at a nonzero offset. */
> + acquired = bpf_task_acquire(task->last_wakee);
The comment is incorrect, that would be &task->last_wakee instead,
this is PTR_TO_BTF_ID | PTR_NESTED.
> + if (!acquired)
> + return 0;
> + bpf_task_release(acquired);
> +
> + return 0;
> +}
> +
> [...]
> +
> +static int test_acquire_release(struct task_struct *task)
> +{
> + struct task_struct *acquired;
> +
> + acquired = bpf_task_acquire(task);
Unfortunately a side effect of this change is that now since
PTR_TO_BTF_ID without ref_obj_id is considered trusted, the bpf_ct_*
functions would begin working with tp_btf args. That probably needs to
be fixed so that they reject them (ideally with a failing test case to
make sure it doesn't resurface), probably with a new suffix __ref/or
__owned as added here [0].
Alexei, since you've suggested avoiding adding that suffix, do you see
any other way out here?
It's questionable whether bpf_ct_set_timeout/status should work for CT
not owned by the BPF program.
[0]: https://lore.kernel.org/bpf/dfb859a6b76a9234baa194e795ae89cb7ca5694b.1662383493.git.lorenzo@xxxxxxxxxx
> + if (!acquired) {
> + err = 1;
> + return 0;
> + }
> +
> + bpf_task_release(acquired);
> +
> + return 0;
> +}
> +
> [...]
