Re: [PATCH 2/3] firmware: google: Unregister driver_info on failure and exit in gsmi

From: Greg Kroah-Hartman
Date: Mon Nov 18 2019 - 03:13:46 EST


On Mon, Nov 18, 2019 at 08:59:32AM +0100, patrick.rudolph@xxxxxxxxxxxxx wrote:
> On Sat, 2019-11-16 at 14:40 +0100, Greg Kroah-Hartman wrote:
> > On Fri, Nov 15, 2019 at 02:48:38PM +0100,
> > patrick.rudolph@xxxxxxxxxxxxx wrote:
> > > From: Arthur Heymans <arthur@xxxxxxxxxxxx>
> > >
> > > Fix a bug where the kernel module couldn't be loaded after
> > > unloading,
> > > as the platform driver wasn't released on exit.
> > >
> > > Signed-off-by: Arthur Heymans <arthur@xxxxxxxxxxxx>
> > > ---
> > > drivers/firmware/google/gsmi.c | 6 ++++++
> > > 1 file changed, 6 insertions(+)
> > >
> > > diff --git a/drivers/firmware/google/gsmi.c
> > > b/drivers/firmware/google/gsmi.c
> > > index edaa4e5d84ad..974c769b75cf 100644
> > > --- a/drivers/firmware/google/gsmi.c
> > > +++ b/drivers/firmware/google/gsmi.c
> > > @@ -1016,6 +1016,9 @@ static __init int gsmi_init(void)
> > > dma_pool_destroy(gsmi_dev.dma_pool);
> > > platform_device_unregister(gsmi_dev.pdev);
> > > pr_info("gsmi: failed to load: %d\n", ret);
> > > +#ifdef CONFIG_PM
> > > + platform_driver_unregister(&gsmi_driver_info);
> > > +#endif
> > > return ret;
> > > }
> > >
> > > @@ -1037,6 +1040,9 @@ static void __exit gsmi_exit(void)
> > > gsmi_buf_free(gsmi_dev.name_buf);
> > > dma_pool_destroy(gsmi_dev.dma_pool);
> > > platform_device_unregister(gsmi_dev.pdev);
> > > +#ifdef CONFIG_PM
> > > + platform_driver_unregister(&gsmi_driver_info);
> >
> > Why the #ifdef here? Why does PM change things?
> >
> The driver is only registered if CONFIG_PM is selected, thus it only
> needs to be unregistered if CONFIG_PM is selected.
>
> See 8942b2d5094b0 for reference.

That is a "fun" abuse of the platform driver interface :(

Why not just have this registration of your device for the "normal"
device your driver binds to? Why create a special platform device
instead? That means you have double the number of "devices" for your
single real device.

thanks,

greg k-h