On Wed, Nov 6, 2019 at 11:27 PM Guillaume Tucker <guillaume.tucker@xxxxxxxxxxxxx> wrote: > On 06/11/2019 22:15, kernelci.org bot wrote: > > +static const struct of_device_id bcm_iproc_gpio_of_match[] __initconst = { > > + { .compatible = "brcm,iproc-gpio-cca" }, > > + {} > > +}; > > +MODULE_DEVICE_TABLE(of, bcm_iproc_gpio_of_match); > > + > > +static struct platform_driver bcm_iproc_gpio_driver = { > > + .driver = { > > + .name = "iproc-xgs-gpio", > > + .owner = THIS_MODULE, > > + .of_match_table = bcm_iproc_gpio_of_match, > > + }, > > + .probe = iproc_gpio_probe, > > + .remove = iproc_gpio_remove, > > +}; > > There's a fix for this which Mark sent yesterday[1] and should > have now been applied, by removing __initconst for the > of_device_id table. So this regression should not be present in > the next linux-next tag. Confirmed, I pushed a new for-next branch with this fix in it. Yours, Linus Walleij