Skip to contentSkip to conversation

gpio: sim: Remove intermediate pointer variable and harden function

PATCH In review Maxwell Doose 2 participants 5 messages Apr 23 – Apr 24 GPIO

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 gpio: sim: Remove intermediate pointer variable and harden function

Remove the *pdev intermediate variable and directly dereference the pointer. While at it, replace sprintf() calls with sysfs_emit() to harden the driver.

---
 drivers/gpio/gpio-sim.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)
drivers/gpio/gpio-sim.c +3 -5
index 13b87c8e6d0c..3c230f94eea2 100644
@@ -691,15 +691,13 @@ static ssize_t gpio_sim_device_config_dev_name_show(struct config_item *item,
 						    char *page)
 {
 	struct gpio_sim_device *dev = to_gpio_sim_device(item);
-	struct platform_device *pdev;
 
 	guard(mutex)(&dev->lock);
 
-	pdev = dev->probe_data.pdev;
-	if (pdev)
-		return sprintf(page, "%s\n", dev_name(&pdev->dev));
+	if (dev->probe_data.pdev)
+		return sysfs_emit(page, "%s\n", dev_name(&dev->probe_data.pdev->dev));
 
-	return sprintf(page, "gpio-sim.%d\n", dev->id);
+	return sysfs_emit(page, "gpio-sim.%d\n", dev->id);
 }
 
 CONFIGFS_ATTR_RO(gpio_sim_device_config_, dev_name);
Signature · 1 line
2.53.0
#1 lore raw
Quoted 17 lines · Maxwell Doose

Remove the *pdev intermediate variable and directly dereference the pointer. While at it, replace sprintf() calls with sysfs_emit() to harden the driver.

---
 drivers/gpio/gpio-sim.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)
drivers/gpio/gpio-sim.c +0 -0
index 13b87c8e6d0c..3c230f94eea2 100644
@@ -691,15 +691,13 @@ static ssize_t gpio_sim_device_config_dev_name_show(struct config_item *item,
                                                    char *page)
 {

drivers/gpio/gpio-sim.c static ssize_t gpio_sim_device_config_dev_name_show(struct config_item *item, · line 698

Quoted 10 lines
drivers/gpio/gpio-sim.c +2 -4
        struct gpio_sim_device *dev = to_gpio_sim_device(item);
-       struct platform_device *pdev;
 
        guard(mutex)(&dev->lock);
 
-       pdev = dev->probe_data.pdev;
-       if (pdev)
-               return sprintf(page, "%s\n", dev_name(&pdev->dev));
+       if (dev->probe_data.pdev)
+               return sysfs_emit(page, "%s\n", dev_name(&dev->probe_data.pdev->dev));

Do you believe the code looks better with more layered dereferences?

drivers/gpio/gpio-sim.c static ssize_t gpio_sim_device_config_dev_name_show(struct config_item *item, · line 700

Quoted 3 lines
drivers/gpio/gpio-sim.c +1 -1
-       return sprintf(page, "gpio-sim.%d\n", dev->id);
+       return sysfs_emit(page, "gpio-sim.%d\n", dev->id);

sprintf() is safe here as we cannot possibly exceed PAGE_SIZE with this format but if you really want to do this, than please send a separate patch converting all configfs show callbacks in the driver.

Bart

Quoted 6 lines
drivers/gpio/gpio-sim.c +0 -1
 }
 
 CONFIGFS_ATTR_RO(gpio_sim_device_config_, dev_name);
--

2.53.0

#2 lore raw
Quoted 3 lines · Bartosz Golaszewski

Do you believe the code looks better with more layered dereferences?

I think some people might value the more explicit dereferencing, and with this we also won't need to handle the overhead of assigning another pointer, but if you want this removed thats fine by me.

Quoted 3 lines

sprintf() is safe here as we cannot possibly exceed PAGE_SIZE with this format but if you really want to do this, than please send a separate patch converting all configfs show callbacks in the driver.

I can do that, but I'd prefer to separate the functional changes with this one, so
expect another patch that fixes these. I'll remove those sysfs_emit changes for now so that the commit history will be clean for when I come back with this second patch.

best regards,
maxwell

#3 lore raw
Quoted 10 lines · Maxwell Doose
Quoted 3 lines · Bartosz Golaszewski

Do you believe the code looks better with more layered dereferences?

I think some people might value the more explicit dereferencing, and with this we also won't need to handle the overhead of assigning another pointer, but if you want this removed thats fine by me.

I prefer to keep it as is.

Quoted 10 lines
Quoted 3 lines

sprintf() is safe here as we cannot possibly exceed PAGE_SIZE with this format but if you really want to do this, than please send a separate patch converting all configfs show callbacks in the driver.

I can do that, but I'd prefer to separate the functional changes with this one, so
expect another patch that fixes these. I'll remove those sysfs_emit changes for now so that the commit history will be clean for when I come back with this second patch.

Sure, let's drop the first part and do the conversion of the configfs show callbacks

Bart

#4 lore raw
Quoted 17 lines · Bartosz Golaszewski
Quoted 4 lines · Maxwell Doose

I think some people might value the more explicit dereferencing, and with this we also won't need to handle the overhead of assigning another pointer, but if you want this removed thats fine by me.

I prefer to keep it as is.

Quoted 7 lines

I can do that, but I'd prefer to separate the functional changes with this one, so
expect another patch that fixes these. I'll remove those sysfs_emit changes for now so that the commit history will be clean for when I come back with this second patch.

Quoted 4 lines

Sure, let's drop the first part and do the conversion of the configfs show callbacks

Bart

Sounds good. I'll get that patch out this afternoon, otherwise I'll be away for most
of the weekend and I'll get it out Sunday evening.

best regards,
maxwell