Re: [PATCH] dca: fix provider device memory leak on domain allocation failure

[Guangshuo Li]

Hi Markus,

Thanks for the suggestion.

On Mon, 21 Sept 2026 at 19:30, Markus Elfring <Markus.Elfring@xxxxxx> wrote:
>
> …
> > +++ b/drivers/dca/dca-core.c
> > @@ -368,8 +368,11 @@ int register_dca_provider(struct dca_provider *dca, struct device *dev)
> >               raw_spin_unlock_irqrestore(&dca_lock, flags);
> >               rc = dca_pci_rc_from_dev(dev);
> >               newdomain = dca_allocate_domain(rc);
> > -             if (!newdomain)
> > +             if (!newdomain) {
> > +                     dca_sysfs_remove_provider(dca);
> >                       return -ENODEV;
> > +             }
> > +
> >               raw_spin_lock_irqsave(&dca_lock, flags);
> …
>
> Please avoid a bit of duplicate source code in this function implementation
> by using another goto chain.
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/coding-style.rst?h=v7.3-rc4#n572
>
> https://elixir.bootlin.com/linux/v7.3-rc3/source/drivers/dca/dca-core.c#L335-L389
>
> Regards,
> Markus

Do you mean using a common error path like this?

        if (dca_providers_blocked) {
                raw_spin_unlock_irqrestore(&dca_lock, flags);
-               dca_sysfs_remove_provider(dca);
                unregister_dca_providers();
-               return -ENODEV;
+               goto err_remove_provider;
        }

        ...

        newdomain = dca_allocate_domain(rc);
        if (!newdomain)
-               return -ENODEV;
+               goto err_remove_provider;

        ...

        kfree(newdomain);
        return 0;
+
+err_remove_provider:
+       dca_sysfs_remove_provider(dca);
+       return -ENODEV;
 }

Would this be the preferred change? If so, I'll follow Krzysztof's
suggestion and include it in a properly organized series for v2.

Thanks,
Guangshuo