Skip to contentSkip to conversation

drivers/media/video/cpia2/cpia2_usb.c: Free previously allocated memory (in array elements) if kmalloc() returns NULL.

PATCH In review Amit Choudhary 3 participants 5 messages Dec 7, 2006 – Dec 8, 2006 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drivers/media/video/cpia2/cpia2_usb.c: Free previously allocated memory (in array elements) if kmalloc() returns NULL.

Description: Free previously allocated memory (in array elements) if kmalloc() returns NULL, in function submit_urbs(), in file drivers/media/video/cpia2/cpia2_usb.c. If the system is low on memory, then previously allocated memory in the same array should be freed up to help the system recover.

drivers/media/video/cpia2/cpia2_usb.c +4 -0
index 28dc6a1..c938638 100644
@@ -640,6 +640,10 @@ static int submit_urbs(struct camera_dat
 		cam->sbuf[i].data =
 		    kmalloc(FRAMES_PER_DESC * FRAME_SIZE_PER_DESC, GFP_KERNEL);
 		if (!cam->sbuf[i].data) {
+			for (--i; i >= 0; i--) {
+				kfree(cam->sbuf[i].data);
+				cam->sbuf[i].data = NULL;
+			}
 			return -ENOMEM;
 		}
 	}
#1 lore raw

Hello,

drivers/media/video/cpia2/cpia2_usb.c static int submit_urbs(struct camera_dat · line 649

Quoted 13 lines
drivers/media/video/cpia2/cpia2_usb.c +4 -0
@@ -640,6 +640,10 @@ static int submit_urbs(struct camera_dat
 		cam->sbuf[i].data =
 		    kmalloc(FRAMES_PER_DESC * FRAME_SIZE_PER_DESC, GFP_KERNEL);
 		if (!cam->sbuf[i].data) {
+			for (--i; i >= 0; i--) {
+				kfree(cam->sbuf[i].data);
+				cam->sbuf[i].data = NULL;
+			}
 			return -ENOMEM;
 		}
 	}

Just for future. Shorter and more readable version of your for(...) thing:

	while (i--) {
		...
	}
Signature · 3 lines
Regards,

	Mariusz Kozlowski
#2 lore raw
Quoted 16 lines · Mariusz Kozlowski

drivers/media/video/cpia2/cpia2_usb.c static int submit_urbs(struct camera_dat · line 649

Quoted 13 lines
drivers/media/video/cpia2/cpia2_usb.c +4 -0
@@ -640,6 +640,10 @@ static int submit_urbs(struct camera_dat
 		cam->sbuf[i].data =
 		    kmalloc(FRAMES_PER_DESC * FRAME_SIZE_PER_DESC, GFP_KERNEL);
 		if (!cam->sbuf[i].data) {
+			for (--i; i >= 0; i--) {
+				kfree(cam->sbuf[i].data);
+				cam->sbuf[i].data = NULL;
+			}
 			return -ENOMEM;
 		}
 	}

Just for future. Shorter and more readable version of your for(...) thing:

Quoted 4 lines
	while (i--) {
		...
	}

No, that is not equivalent.

You want

	while (i-- >= 0) {
		...
	}
#3 lore raw

Hello,

Quoted 7 lines
Quoted 6 lines

Just for future. Shorter and more readable version of your for(...) thing:

	while (i--) {
		...
	}
Quoted 7 lines

No, that is not equivalent.

You want

	while (i-- >= 0) {
		...
	}

Not really. That will stop at -1 not 0.

Signature · 3 lines
Regards,

	Mariusz Kozlowski
#4 lore raw
Quoted 9 lines · Mariusz Kozlowski
Quoted 9 lines
Quoted 6 lines

Just for future. Shorter and more readable version of your for(...) thing:

	while (i--) {
		...
	}

No, that is not equivalent.

Quoted 8 lines
Quoted 5 lines

You want

	while (i-- >= 0) {
		...
	}

Not really. That will stop at -1 not 0.

It depends on your intent. Generally speaking,

	while (i--) {
		...
	}

is never what you want. Adding the check for being greater than 0 stops potential bugs from signed int i being negative. The only drawback on x86 is that it sets a byte based on the greater condition with setg and tests it later with testb for every iteration. This use of testb will _always_ use the same addressable byte registers for both its operands.

Based on this particular patch, I agree that

	while (i-- > 0) {
		...
	}

will do the job. This is equivalent to the original for loop and ensures that a negative value of i is never iterated on (even though it admittedly would never be negative in this instance to begin with).

David