On Mon, Jul 17, 2023 at 05:40:11PM +0300, Andy Shevchenko wrote:
> On Mon, Jul 17, 2023 at 04:27:43PM +0200, Uwe Kleine-König wrote:
>
> > Note that irq_domain_remove() also isn't called so there is another
> > resource leak introduced by 812d47889a8e ("gpio/mvebu: Use
> > irq_domain_add_linear")
>
> I believe it's only theoretical issue as quite likely GPIO will be used and
> hence one can't unbound it. Or unbound is unconditional? In that case we
> have much bigger issue in the kernel.
You might want to test
echo f1018100.gpio > /sys/bus/platform/drivers/mvebu-gpio/unbind
(where f1018100.gpio is the name of one of links in
/sys/bus/platform/drivers/mvebu-gpio#) and if that succeeds you should
consider
diff --git a/drivers/gpio/gpio-mvebu.c b/drivers/gpio/gpio-mvebu.c
index a68f682aec01..620ab247abff 100644
--- a/drivers/gpio/gpio-mvebu.c
+++ b/drivers/gpio/gpio-mvebu.c
@@ -1306,6 +1306,7 @@ static struct platform_driver mvebu_gpio_driver = {
.driver = {
.name = "mvebu-gpio",
.of_match_table = mvebu_gpio_of_match,
+ .suppress_bind_attrs = true,
},
.probe = mvebu_gpio_probe,
.suspend = mvebu_gpio_suspend,
which prevents that path and so the leak (by not providing
/sys/bus/platform/drivers/mvebu-gpio/unbind).
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | https://www.pengutronix.de/ |
Attachment:
signature.asc
Description: PGP signature
