Skip to contentSkip to conversation

gpiolib: check for overflow when reading the 'ngpios' property

PATCH In review Bartosz Golaszewski 3 participants 6 messages Mar 6, 2022 – Mar 7, 2022 GPIO

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 gpiolib: check for overflow when reading the 'ngpios' property

The ngpio fields both in struct gpio_device as well as gpio_chip are 16-bit unsigned integers. Let's not risk an overflow and check if the property value represented as a 32-bit unsigned integer is not greater than U16_MAX.

Fixes 9dbd1ab20509 ("gpiolib: check the 'ngpios' property in core gpiolib code")
---
 drivers/gpio/gpiolib.c | 5 +++++
 1 file changed, 5 insertions(+)
drivers/gpio/gpiolib.c +5 -0
index a3d14277f17c..3c4f47b9ab57 100644
@@ -677,6 +677,11 @@ int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data,
 		else if (ret)
 			goto err_free_descs;
 
+		if (ngpios > U16_MAX) {
+			ret = EINVAL;
+			goto err_free_descs;
+		}
+
 		gc->ngpio = ngpios;
 	}
 
Signature · 1 line
2.30.1
#1 lore raw
Quoted 5 lines · Bartosz Golaszewski

The ngpio fields both in struct gpio_device as well as gpio_chip are 16-bit unsigned integers. Let's not risk an overflow and check if the property value represented as a 32-bit unsigned integer is not greater than U16_MAX.

...

drivers/gpio/gpiolib.c int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data, · line 683

Quoted 4 lines
drivers/gpio/gpiolib.c +4 -0
+               if (ngpios > U16_MAX) {
+                       ret = EINVAL;
+                       goto err_free_descs;
+               }

I don't think it's a fatal error in this case. I would perhaps print a warning and simply use a masked (which is done implicitly by an assignment of the different type) value. Note, the above is buggy on the buggy DTs, where the upper part of the value is not used. After this patch you effectively make a regression on, yes, broken DTs.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#2 lore raw
Quoted 14 lines · Andy Shevchenko
Quoted 5 lines · Bartosz Golaszewski

The ngpio fields both in struct gpio_device as well as gpio_chip are 16-bit unsigned integers. Let's not risk an overflow and check if the property value represented as a 32-bit unsigned integer is not greater than U16_MAX.

...

drivers/gpio/gpiolib.c int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data, · line 683

Quoted 4 lines
drivers/gpio/gpiolib.c +4 -0
+               if (ngpios > U16_MAX) {
+                       ret = EINVAL;
+                       goto err_free_descs;
+               }
Quoted 5 lines

I don't think it's a fatal error in this case. I would perhaps print a warning and simply use a masked (which is done implicitly by an assignment of the different type) value. Note, the above is buggy on the buggy DTs, where the upper part of the value is not used. After this patch you effectively make a regression on, yes, broken DTs.

Like

if (ngpios > U16_MAX)
chip_warn(gc, "line cnt %u is greater than supported; use
%u\n", ngpios, (u16)ngpio);

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
Quoted 22 lines · Andy Shevchenko
Quoted 14 lines · Andy Shevchenko
Quoted 5 lines · Bartosz Golaszewski

The ngpio fields both in struct gpio_device as well as gpio_chip are 16-bit unsigned integers. Let's not risk an overflow and check if the property value represented as a 32-bit unsigned integer is not greater than U16_MAX.

...

drivers/gpio/gpiolib.c int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data, · line 683

Quoted 4 lines
drivers/gpio/gpiolib.c +4 -0
+               if (ngpios > U16_MAX) {
+                       ret = EINVAL;
+                       goto err_free_descs;
+               }
Quoted 5 lines

I don't think it's a fatal error in this case. I would perhaps print a warning and simply use a masked (which is done implicitly by an assignment of the different type) value. Note, the above is buggy on the buggy DTs, where the upper part of the value is not used. After this patch you effectively make a regression on, yes, broken DTs.

Quoted 5 lines

Like

if (ngpios > U16_MAX)
chip_warn(gc, "line cnt %u is greater than supported; use
%u\n", ngpios, (u16)ngpio);

Or to be on safer side move this after == 0 check as

	if (gc->ngpio != ngpios)
		chip_warn(gc, "line cnt %u is greater than supported; use %u\n", ngpios, gc->ngpio);
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw
Quoted 26 lines · Andy Shevchenko
Quoted 24 lines · Andy Shevchenko
Quoted 14 lines · Andy Shevchenko
Quoted 5 lines · Bartosz Golaszewski

The ngpio fields both in struct gpio_device as well as gpio_chip are 16-bit unsigned integers. Let's not risk an overflow and check if the property value represented as a 32-bit unsigned integer is not greater than U16_MAX.

...

drivers/gpio/gpiolib.c int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data, · line 683

Quoted 4 lines
drivers/gpio/gpiolib.c +4 -0
+               if (ngpios > U16_MAX) {
+                       ret = EINVAL;
+                       goto err_free_descs;
+               }
Quoted 5 lines

I don't think it's a fatal error in this case. I would perhaps print a warning and simply use a masked (which is done implicitly by an assignment of the different type) value. Note, the above is buggy on the buggy DTs, where the upper part of the value is not used. After this patch you effectively make a regression on, yes, broken DTs.

Like

Quoted 9 lines
Quoted 3 lines

if (ngpios > U16_MAX)
chip_warn(gc, "line cnt %u is greater than supported; use
%u\n", ngpios, (u16)ngpio);

Or to be on safer side move this after == 0 check as

        if (gc->ngpio != ngpios)
                chip_warn(gc, "line cnt %u is greater than supported; use %u\n", ngpios, gc->ngpio);

ngpios is not necessarily used so this check must be in the scope of the device property read (inside the if (gc->ngpio == 0) { block).

Bart

Quoted 5 lines

--
With Best Regards,
Andy Shevchenko

#5 lore raw
Quoted 31 lines · Bartosz Golaszewski
Quoted 29 lines · Andy Shevchenko
Quoted 27 lines · Andy Shevchenko
Quoted 14 lines · Andy Shevchenko
Quoted 4 lines · Bartosz Golaszewski

The ngpio fields both in struct gpio_device as well as gpio_chip are 16-bit unsigned integers. Let's not risk an overflow and check if the property value represented as a 32-bit unsigned integer is not greater than U16_MAX.

...

drivers/gpio/gpiolib.c int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data, · line 683

Quoted 4 lines
drivers/gpio/gpiolib.c +4 -0
+               if (ngpios > U16_MAX) {
+                       ret = EINVAL;
+                       goto err_free_descs;
+               }
Quoted 5 lines

I don't think it's a fatal error in this case. I would perhaps print a warning and simply use a masked (which is done implicitly by an assignment of the different type) value. Note, the above is buggy on the buggy DTs, where the upper part of the value is not used. After this patch you effectively make a regression on, yes, broken DTs.

Like

if (ngpios > U16_MAX)
chip_warn(gc, "line cnt %u is greater than supported; use
%u\n", ngpios, (u16)ngpio);

Quoted 8 lines
Quoted 5 lines

Or to be on safer side move this after == 0 check as

        if (gc->ngpio != ngpios)
                chip_warn(gc, "line cnt %u is greater than supported; use %u\n", ngpios, gc->ngpio);

ngpios is not necessarily used so this check must be in the scope of the device property read (inside the if (gc->ngpio == 0) { block).

Can be done as

        if (gc->ngpio == 0) {
		...
	} else {
		ngpios = gc->ngpio;
	}

        if (gc->ngpio == 0) {
		...
	}

	if (gc->ngpio != ngpios)
		chip_warn(gc, "line cnt %u is greater than supported; use %u\n", ngpios, gc->ngpio);

The point of this exercise is to avoid hard coded type of the variable in a few places, so if gc->ngpio and/or ngpios have changed type in the future, you don't need to change this code.

Signature · 2 lines
With Best Regards,
Andy Shevchenko