On Tue, Jun 02, 2026 at 01:31:44PM +0300, Svyatoslav Ryhel wrote:

вт, 2 черв. 2026 р. о 11:24 Andy Shevchenko [off-list ref] пише:

quoted

On Mon, Jun 01, 2026 at 06:18:25PM +0300, Svyatoslav Ryhel wrote:

...

quoted

quoted

+     device_for_each_child_node_scoped(lm3533->dev, child) {

quoted

+             if (!fwnode_device_is_available(child))
+                     continue;

Do we need this check?

This is nice to have if the node is disabled. If we assume that there
are no disabled nodes, I can remove it.

It's already implied. See

static struct fwnode_handle *
of_fwnode_get_next_child_node(const struct fwnode_handle *fwnode, struct fwnode_handle *child)
{
        return of_fwnode_handle(of_get_next_available_child(to_of_node(fwnode), to_of_node(child)));
}

And I believe it's written somewhere in the documentation (if not, feel free to
patch that).

...

quoted

quoted

+     ret = sysfs_create_group(&dev->kobj, &lm3533_attribute_group);

No way. You should use .dev_groups.

I did not change how driver does this, just swapped lm3533->dev to
dev. I will set is back as it was.

This is a serious race condition that needs to be addressed. Since you are
touching this driver the fixes against known issues probably are the first
things that have to be done.

quoted

quoted

+     if (ret) {
+             dev_err(dev, "failed to create sysfs attributes\n");
              goto err_unregister;
      }

...

quoted

Can you think on how to split this change to smaller steps? I believe it's
possible.

No, I am done with tinkering with this patchset. It is broken enough
and it has inflated enough.

Probably you don't want this to be reviewed then? I believe other kernel
developers and maintainers will ask you the same.

-- 
With Best Regards,
Andy Shevchenko