Skip to contentSkip to conversation

media: stkwebcam: move stk_camera_read_reg() scratch buffer to struct stk_camera

PATCH In review [email protected] 3 participants 8 messages Mar 12, 2022 – Apr 27, 2022 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 media: stkwebcam: move stk_camera_read_reg() scratch buffer to struct stk_camera

From: Tom Rix <[email protected]>

In stk_camera_read_reg() a single byte buffer is alloc-ed and freed on every function call. Since the size is known, move the buffer to the struct stk_camera where it will be alloc-ed and freed once.

---
 drivers/media/usb/stkwebcam/stk-webcam.c | 11 ++---------
 drivers/media/usb/stkwebcam/stk-webcam.h |  2 ++
 2 files changed, 4 insertions(+), 9 deletions(-)
drivers/media/usb/stkwebcam/stk-webcam.c +2 -9
index 5b822214ccc5c..787edb3d47c23 100644
@@ -150,25 +150,18 @@ int stk_camera_write_reg(struct stk_camera *dev, u16 index, u8 value)
 int stk_camera_read_reg(struct stk_camera *dev, u16 index, u8 *value)
 {
 	struct usb_device *udev = dev->udev;
-	unsigned char *buf;
 	int ret;
 
-	buf = kmalloc(sizeof(u8), GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-
 	ret = usb_control_msg(udev, usb_rcvctrlpipe(udev, 0),
 			0x00,
 			USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
 			0x00,
 			index,
-			buf,
+			&dev->read_reg_scratch,
 			sizeof(u8),
 			500);
 	if (ret >= 0)
-		*value = *buf;
-
-	kfree(buf);
+		*value = dev->read_reg_scratch;
 
 	if (ret < 0)
 		return ret;
drivers/media/usb/stkwebcam/stk-webcam.h +2 -0
index 14519e5308b18..136decffe9ced 100644
@@ -105,6 +105,8 @@ struct stk_camera {
 	struct list_head sio_avail;
 	struct list_head sio_full;
 	unsigned sequence;
+
+	u8 read_reg_scratch;
 };
 
 #define vdev_to_camera(d) container_of(d, struct stk_camera, vdev)
Signature · 1 line
2.26.3
#1 lore raw

Hi Trix,

Quoted 27 lines · On 3/12/22 20:30, [email protected] wrote:

From: Tom Rix <[email protected]>

In stk_camera_read_reg() a single byte buffer is alloc-ed and freed on every function call. Since the size is known, move the buffer to the struct stk_camera where it will be alloc-ed and freed once.

---
  drivers/media/usb/stkwebcam/stk-webcam.c | 11 ++---------
  drivers/media/usb/stkwebcam/stk-webcam.h |  2 ++
  2 files changed, 4 insertions(+), 9 deletions(-)
drivers/media/usb/stkwebcam/stk-webcam.c +0 -4
index 5b822214ccc5c..787edb3d47c23 100644
@@ -150,25 +150,18 @@ int stk_camera_write_reg(struct stk_camera *dev, u16 index, u8 value)
  int stk_camera_read_reg(struct stk_camera *dev, u16 index, u8 *value)
  {
  	struct usb_device *udev = dev->udev;
-	unsigned char *buf;
  	int ret;
 
-	buf = kmalloc(sizeof(u8), GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;

drivers/media/usb/stkwebcam/stk-webcam.c int stk_camera_write_reg(struct stk_camera *dev, u16 index, u8 value) · line 162

Quoted 10 lines
drivers/media/usb/stkwebcam/stk-webcam.c +1 -2
-
  	ret = usb_control_msg(udev, usb_rcvctrlpipe(udev, 0),
  			0x00,
  			USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
  			0x00,
  			index,
-			buf,
+			&dev->read_reg_scratch,
  			sizeof(u8),
  			500);

Wouldn't it be better to move to modern USB API like usb_control_msg_recv()?

This API does not require buffer to be allocated via kmalloc(), so you will be able to use value directly.

With regards,
Pavel Skripkin

#2 lore raw
In reply to [email protected]
Quoted 18 lines · On 3/12/22 20:30, [email protected] wrote:

From: Tom Rix <[email protected]>

In stk_camera_read_reg() a single byte buffer is alloc-ed and freed on every function call. Since the size is known, move the buffer to the struct stk_camera where it will be alloc-ed and freed once.

---
  drivers/media/usb/stkwebcam/stk-webcam.c | 11 ++---------
  drivers/media/usb/stkwebcam/stk-webcam.h |  2 ++
  2 files changed, 4 insertions(+), 9 deletions(-)
drivers/media/usb/stkwebcam/stk-webcam.c +0 -0
index 5b822214ccc5c..787edb3d47c23 100644

drivers/media/usb/stkwebcam/stk-webcam.c int stk_camera_write_reg(struct stk_camera *dev, u16 index, u8 value) · line 150

Quoted 1 line
drivers/media/usb/stkwebcam/stk-webcam.c +0 -0
@@ -150,25 +150,18 @@ int stk_camera_write_reg(struct stk_camera *dev, u16 index, u8 value)
  int stk_camera_read_reg(struct stk_camera *dev, u16 index, u8 *value)

And just random note: there are 4 possible uninit value bugs.

stk_start_stream() calls stk_camera_read_reg 4 times, but ignores return values.

stk_camera_read_reg() should have __must_check annotation and return value should be checked on each call.

If you have time you can take care of it :) Or I will fix it one day...

With regards,
Pavel Skripkin

author
#3 lore raw
Quoted 24 lines · Pavel Skripkin
Quoted 18 lines · On 3/12/22 20:30, [email protected] wrote:

From: Tom Rix <[email protected]>

In stk_camera_read_reg() a single byte buffer is alloc-ed and freed on every function call.  Since the size is known, move the buffer to the struct stk_camera where it will be alloc-ed and freed once.

---
  drivers/media/usb/stkwebcam/stk-webcam.c | 11 ++--------- drivers/media/usb/stkwebcam/stk-webcam.h |  2 ++

  2 files changed, 4 insertions(+), 9 deletions(-)
drivers/media/usb/stkwebcam/stk-webcam.c +0 -0
index 5b822214ccc5c..787edb3d47c23 100644

drivers/media/usb/stkwebcam/stk-webcam.c int stk_camera_write_reg(struct stk_camera · line 151

Quoted 2 lines
drivers/media/usb/stkwebcam/stk-webcam.c +0 -0
@@ -150,25 +150,18 @@ int stk_camera_write_reg(struct stk_camera 
 *dev, u16 index, u8 value)
   int stk_camera_read_reg(struct stk_camera *dev, u16 index, u8 *value)

And just random note: there are 4 possible uninit value bugs.

Quoted 6 lines

stk_start_stream() calls stk_camera_read_reg 4 times, but ignores return values.

stk_camera_read_reg() should have __must_check annotation and return value should be checked on each call.

If you have time you can take care of it :) Or I will fix it one day...

These do show up in my usual static analysis and it why I was looking at this file.

And was sidetracked by the short malloc.

Unfortunately I looked and there are many other similar instances treewide ~100

These aren't caught in checkpatch, so working on that..

Tom

Quoted 8 lines

With regards,
Pavel Skripkin

#4 lore raw

Hi Tom,

Quoted 8 lines · Tom Rix

These do show up in my usual static analysis and it why I was looking at this file.

And was sidetracked by the short malloc.

Unfortunately I looked and there are many other similar instances treewide ~100

Most of them are in very old drivers and I don't think they ever be fixed. I've looked into one bug reported by syzkaller and there was like 30 calls w/o proper error handling in one driver. Redoing whole driver logic without access to hw seems dangerous :))

Quoted 2 lines

These aren't caught in checkpatch, so working on that..

I think, it's not checkpath responsibility. Maybe it worth adding such check to smatch. I tried to implement such checker, but never finished it :(

With regards,
Pavel Skripkin

author
#5 lore raw
Quoted 6 lines · Pavel Skripkin

Hi Tom,

Quoted 3 lines · Tom Rix

These do show up in my usual static analysis and it why I was looking at this file.

Quoted 7 lines
Quoted 5 lines

And was sidetracked by the short malloc.

Unfortunately I looked and there are many other similar instances treewide ~100

Most of them are in very old drivers and I don't think they ever be fixed. I've looked into one bug reported by syzkaller and there was like 30 calls w/o proper error handling in one driver. Redoing whole driver logic without access to hw seems dangerous :))

From the checkpatch change below, there are about 150 dinky allocs treewide

Here is a refactoring

https://lore.kernel.org/lkml/[email protected]/

Quoted 8 lines
Quoted 4 lines

These aren't caught in checkpatch, so working on that..

I think, it's not checkpath responsibility. Maybe it worth adding such check to smatch. I tried to implement such checker, but never finished it :(

Poking new development to not do dinky allocs I think is worth it, here is my checkpatch patch

https://lore.kernel.org/lkml/[email protected]/

steal the regex for smatch.

Tom

Quoted 4 lines

With regards,
Pavel Skripkin

#6 lore raw
In reply to [email protected]

Hi Tom,

Quoted 6 lines · On 12/03/2022 18:30, [email protected] wrote:

From: Tom Rix <[email protected]>

In stk_camera_read_reg() a single byte buffer is alloc-ed and freed on every function call. Since the size is known, move the buffer to the struct stk_camera where it will be alloc-ed and freed once.

I read the replies to this patch, but I am not certain if you still want this patch to be merged, or will make a v2. I have no problem applying this patch as-is, but I just want to have confirmation that there won't be a v2.

Regards,

Hans

Quoted 52 lines
---
 drivers/media/usb/stkwebcam/stk-webcam.c | 11 ++---------
 drivers/media/usb/stkwebcam/stk-webcam.h |  2 ++
 2 files changed, 4 insertions(+), 9 deletions(-)
drivers/media/usb/stkwebcam/stk-webcam.c +2 -9
index 5b822214ccc5c..787edb3d47c23 100644
@@ -150,25 +150,18 @@ int stk_camera_write_reg(struct stk_camera *dev, u16 index, u8 value)
 int stk_camera_read_reg(struct stk_camera *dev, u16 index, u8 *value)
 {
 	struct usb_device *udev = dev->udev;
-	unsigned char *buf;
 	int ret;
 
-	buf = kmalloc(sizeof(u8), GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-
 	ret = usb_control_msg(udev, usb_rcvctrlpipe(udev, 0),
 			0x00,
 			USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
 			0x00,
 			index,
-			buf,
+			&dev->read_reg_scratch,
 			sizeof(u8),
 			500);
 	if (ret >= 0)
-		*value = *buf;
-
-	kfree(buf);
+		*value = dev->read_reg_scratch;
 
 	if (ret < 0)
 		return ret;
drivers/media/usb/stkwebcam/stk-webcam.h +2 -0
index 14519e5308b18..136decffe9ced 100644
@@ -105,6 +105,8 @@ struct stk_camera {
 	struct list_head sio_avail;
 	struct list_head sio_full;
 	unsigned sequence;
+
+	u8 read_reg_scratch;
 };
 
 #define vdev_to_camera(d) container_of(d, struct stk_camera, vdev)
author
#7 lore raw
Quoted 5 lines · Hans Verkuil

Hi Tom,

Quoted 2 lines · On 12/03/2022 18:30, [email protected] wrote:

From: Tom Rix <[email protected]>

Quoted 7 lines
Quoted 4 lines

In stk_camera_read_reg() a single byte buffer is alloc-ed and freed on every function call. Since the size is known, move the buffer to the struct stk_camera where it will be alloc-ed and freed once.

I read the replies to this patch, but I am not certain if you still want this patch to be merged, or will make a v2. I have no problem applying this patch as-is, but I just want to have confirmation that there won't be a v2.

Merged as-is is fine.

Thanks

Tom

Quoted 56 lines

Regards,

Hans

Quoted 51 lines
---
  drivers/media/usb/stkwebcam/stk-webcam.c | 11 ++---------
  drivers/media/usb/stkwebcam/stk-webcam.h |  2 ++
  2 files changed, 4 insertions(+), 9 deletions(-)
drivers/media/usb/stkwebcam/stk-webcam.c +2 -9
index 5b822214ccc5c..787edb3d47c23 100644
@@ -150,25 +150,18 @@ int stk_camera_write_reg(struct stk_camera *dev, u16 index, u8 value)
  int stk_camera_read_reg(struct stk_camera *dev, u16 index, u8 *value)
  {
  	struct usb_device *udev = dev->udev;
-	unsigned char *buf;
  	int ret;
 
-	buf = kmalloc(sizeof(u8), GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-
  	ret = usb_control_msg(udev, usb_rcvctrlpipe(udev, 0),
  			0x00,
  			USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
  			0x00,
  			index,
-			buf,
+			&dev->read_reg_scratch,
  			sizeof(u8),
  			500);
  	if (ret >= 0)
-		*value = *buf;
-
-	kfree(buf);
+		*value = dev->read_reg_scratch;
 
  	if (ret < 0)
  		return ret;
drivers/media/usb/stkwebcam/stk-webcam.h +2 -0
index 14519e5308b18..136decffe9ced 100644
@@ -105,6 +105,8 @@ struct stk_camera {
  	struct list_head sio_avail;
  	struct list_head sio_full;
  	unsigned sequence;
+
+	u8 read_reg_scratch;
  };
 
  #define vdev_to_camera(d) container_of(d, struct stk_camera, vdev)