Re: [PATCH net v3 1/3] macsec: introduce dedicated workqueue for SA crypto cleanup

[Jinliang Zheng]

On Sun, 11 May 2026, Sabrina Dubroca wrote:
> > +	macsec_wq = alloc_ordered_workqueue("macsec", 0);
>
> Why did you pick an ordered workqueue instead of a normal one?

I chose an ordered workqueue for its serialized execution, which
avoids potential ordering issues between concurrent SA destructions.
In hindsight, there is no actual ordering dependency between SA
cleanups, so a normal workqueue is sufficient here.

Should I send a v4 with that switched to alloc_workqueue()?

Thanks,
Jinliang

> > +err_destroy_wq:
> > +	/* Precautionary, mirrors macsec_exit() to stay safe if work
> > +	 * ever becomes queueable before this point in the future.
> > +	 */
>
> I don't think this comment is really necessary, but ok.
>
> (I missed the discussion on v2, it's not possible to create SAs at
> this point, but it's reasonable to add the rcu_barrier here)