On Fri, Apr 25, 2025 at 08:23:41PM +0200, Christoph Hellwig wrote:
> On Thu, Apr 24, 2025 at 11:21:35PM +0800, Ming Lei wrote:
> > +struct elv_change_ctx {
> > + const char *name;
> > + bool uevent;
>
> There's only one caller that wants to supress the uevents. So maybe
> invert the polarity so that it only has to be set in one place,
> which also documents how setting the initial scheduler is special a
> bit better.
OK.
>
> > - ret = elv_register_queue(q, true);
> > + ret = elv_register_queue(q, ctx->uevent);
>
> .. and pass the ctx on to elv_register_queue instead of converting
> paramter types. Although that might be woeth doing later when
> another argument derived from ctx gets passed as well.
ctx isn't very useful for elv_register_queue(), which can't figure out
the exact elevator queue to use from `ctx` since there are two(old, new).
Thanks,
Ming
