Re: [PATCH v5 1/3] cgroup/bpf: use a dedicated workqueue for cgroup bpf destruction

[Chen Ridong]

On 2024/9/26 20:49, Michal Koutný wrote:
> Hello.
>
> On Mon, Sep 23, 2024 at 11:43:50AM GMT, Chen Ridong <chenridong@xxxxxxxxxxxxxxx> wrote:
> > +static int __init cgroup_bpf_wq_init(void)
> > +{
> > +	cgroup_bpf_destroy_wq = alloc_workqueue("cgroup_bpf_destroy", 0, 1);
> > +	WARN_ON_ONCE(!cgroup_bpf_destroy_wq);
> > +	return 0;
> > +}
> > +core_initcall(cgroup_bpf_wq_init);
>
> I think hard fail (panic() if you want to avoid BUG_ON) would be
> warranted here and mere warning would leave system exposed to worse
> errors later (and _ONCE in an initcall looks unnecessary).
Thank you. I think panic() is alright.

Best regards,
Ridong

> Maybe look at other global wqs. I see that returning -ENOMEM might be an
> option, however, I don't see that initcall's return value would be
> processed anywhere currently :-/
>
> Besides this allocation failpath this is a sensible change to me.
>
> Thanks,
> Michal