Re: [PATCH v16 02/14] power: reset: reboot-mode: Add device tree node-based registration

[Bartosz Golaszewski <bartosz.golaszewski@linaro.org>]

On Thu, 16 Oct 2025 at 19:19, Shivendra Pratap
<shivendra.pratap@oss.qualcomm.com> wrote:
> >>
> >> -                       info = devm_kzalloc(reboot->dev, sizeof(*info), GFP_KERNEL);
> >
> > This change is good - devres should not be used in subsystem library
> > code, only in drivers - but it doesn't seem to belong here, can you
> > please separate it out and make it backportable?
>
> sure. Just to confirm we should separate out the devm_kzalloc part of the
> change and add a fixes tag.
>

And preferably put it first in the series to avoid conflicts.

> >> @@ -123,8 +136,11 @@ int reboot_mode_register(struct reboot_mode_driver *reboot)
> >>                 return 0;
> >>
> >>  error:
> >> -               list_for_each_entry(info, &reboot->head, list)
> >> +               list_for_each_entry_safe(info, next, &reboot->head, list) {
> >> +                       list_del(&info->list);
> >
> > Same here, not deleting the entries currently seems like a bug? Do we
> > depend on the driver detach to clean up the resources on failure?
>
> sure, so this should also go as fixes? and should we remove the other
> dev_err(printk) also as fixes? or that can still got with the change
> where we add fwnode based registration?
>

It doesn't seem to be strictly required by current code as the users
use it "correctly" but if the API becomes used in different ways - for
instance the structure may be reused after failure - it's a good idea
to backport it. In general we should undo everything we did in the
same function if we fail at some point.

Bart