In case we have a device instantiated via DT or other means than
via new_device sysfs node, the collision with the latter is possible.
Prevent such collisions by allocating user instantiated devices with
higher IDs, currently set to 1024.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/gpio/gpio-aggregator.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/gpio/gpio-aggregator.c b/drivers/gpio/gpio-aggregator.c
index 4a470dd8b75d..8892cb37ad79 100644
--- a/drivers/gpio/gpio-aggregator.c
+++ b/drivers/gpio/gpio-aggregator.c
@@ -26,6 +26,7 @@
 #include <linux/gpio/driver.h>
 #include <linux/gpio/machine.h>
 
+#define AGGREGATOR_MIN_DEVID 1024
 #define AGGREGATOR_MAX_GPIOS 512
 
 /*
@@ -135,7 +136,7 @@ static ssize_t new_device_store(struct device_driver *driver, const char *buf,
 	}
 
 	mutex_lock(&gpio_aggregator_lock);
-	id = idr_alloc(&gpio_aggregator_idr, aggr, 0, 0, GFP_KERNEL);
+	id = idr_alloc(&gpio_aggregator_idr, aggr, AGGREGATOR_MIN_DEVID, 0, GFP_KERNEL);
 	mutex_unlock(&gpio_aggregator_lock);
 
 	if (id < 0) {
-- 
2.40.0.1.gaa8946217a0b

Hi Andy,

Thanks for your patch!

On Thu, Jun 15, 2023 at 3:51 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
> In case we have a device instantiated via DT or other means than
> via new_device sysfs node, the collision with the latter is possible.
> Prevent such collisions by allocating user instantiated devices with
> higher IDs, currently set to 1024.

Can you please elaborate? How exactly is this possible?

Aggregators instantiated through sysfs are named "gpio-aggregator.<n>",
and are IDR-based.
Aggregators instantiated from DT are named "<unit-address>.<node-name>".
How can this conflict? When instantiated from ACPI?
What am I missing?

> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

> --- a/drivers/gpio/gpio-aggregator.c
> +++ b/drivers/gpio/gpio-aggregator.c
> @@ -26,6 +26,7 @@
>  #include <linux/gpio/driver.h>
>  #include <linux/gpio/machine.h>
>
> +#define AGGREGATOR_MIN_DEVID 1024
>  #define AGGREGATOR_MAX_GPIOS 512
>
>  /*
> @@ -135,7 +136,7 @@ static ssize_t new_device_store(struct device_driver *driver, const char *buf,
>         }
>
>         mutex_lock(&gpio_aggregator_lock);
> -       id = idr_alloc(&gpio_aggregator_idr, aggr, 0, 0, GFP_KERNEL);
> +       id = idr_alloc(&gpio_aggregator_idr, aggr, AGGREGATOR_MIN_DEVID, 0, GFP_KERNEL);

Iff this would solve an issue, it would be only temporarily, until someone
instantiates 1024 aggregators through some other means ;-)

>         mutex_unlock(&gpio_aggregator_lock);
>
>         if (id < 0) {

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

On Thu, Jun 15, 2023 at 04:54:14PM +0200, Geert Uytterhoeven wrote:
> On Thu, Jun 15, 2023 at 3:51 PM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > In case we have a device instantiated via DT or other means than
> > via new_device sysfs node, the collision with the latter is possible.
> > Prevent such collisions by allocating user instantiated devices with
> > higher IDs, currently set to 1024.
> 
> Can you please elaborate? How exactly is this possible?
> 
> Aggregators instantiated through sysfs are named "gpio-aggregator.<n>",
> and are IDR-based.
> Aggregators instantiated from DT are named "<unit-address>.<node-name>".
> How can this conflict? When instantiated from ACPI?
> What am I missing?

Nothing. It's me who misunderstood how OF platform device naming schema works.

So this patch can be discarded as we never will have gpio-delay available for
removal via delete_device sysfs node.

Bart, tell me if you need a new version w/o this patch (but note that b4 can
handle this case with

	b4 -slt -P1,2,4,5 ...

).

-- 
With Best Regards,
Andy Shevchenko