On Thu, Jan 19, 2023 at 05:58:32PM -0600, David Vernet wrote:
> +
> +For example:
> +
> +.. code-block:: c
> +
> + /**
> + * A trivial example tracepoint program that shows how to
> + * acquire and release a struct bpf_cpumask *.
> + */
> + SEC("tp_btf/task_newtask")
> + int BPF_PROG(task_acquire_release_example, struct task_struct *task, u64 clone_flags)
> + {
> + struct bpf_cpumask *cpumask, *acquired;
> +
> + cpumask = bpf_cpumask_create();
> + if (!cpumask)
> + return 1;
> +
> + acquired = bpf_cpumask_acquire(cpumask);
> + bpf_cpumask_release(cpumask);
> + bpf_cpumask_acquire(acquired);
> +
> + return 0;
> + }
As the first example in the doc it was... alarming :)
I've read it as it says that bpf_cpumask_acquire has to be called on
freshly created cpumask before it can be used.
I've started to doubt by code reading skills of the previous patches :)
A basic example is probably necessary to introduce the concept.
Or this example should have bpf_cpumask_set_cpu right after create and
more alu ops after release with comments to demonstrate the point.
