[PATCH] staging: goldfish_audio.c: sparse warning of incorrect type

[PATCH] staging: goldfish_audio.c: sparse warning of incorrect type

[Sudip Mukherjee]

fixed sparse warning of incorrect type in argument 1 and incorrect type in argument 2
it was directly dereferencing a __iomem pointer , which will work in x86 but will fail in other architectures.

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
---

hi,
can you please reveiew the patch and check if the approach is correct. If it is correct then I can send another patch to correct the other similar warnings present in this file. 

 drivers/staging/goldfish/goldfish_audio.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/goldfish/goldfish_audio.c b/drivers/staging/goldfish/goldfish_audio.c
index c89d0b8..f9a13e7 100644
--- a/drivers/staging/goldfish/goldfish_audio.c
+++ b/drivers/staging/goldfish/goldfish_audio.c
@@ -118,6 +118,7 @@ static ssize_t goldfish_audio_read(struct file *fp, char __user *buf,
 	struct goldfish_audio *data = fp->private_data;
 	int length;
 	int result = 0;
+	void *buffer;
 
 	if (!data->read_supported)
 		return -ENODEV;
@@ -133,9 +134,15 @@ static ssize_t goldfish_audio_read(struct file *fp, char __user *buf,
 						AUDIO_READ_BUFFER_AVAILABLE);
 
 		/* copy data to user space */
-		if (copy_to_user(buf, data->read_buffer, length))
+		buffer = kzalloc(length, GFP_KERNEL);
+		if (buffer == NULL)
+			return -ENOMEM;
+		memcpy_fromio(buffer, data->read_buffer, length);
+		if (copy_to_user((void __user *)buf, buffer, length)) {
+			kfree(buffer);
 			return -EFAULT;
-
+		}
+		kfree(buffer);
 		result += length;
 		buf += length;
 		count -= length;
-- 
1.8.1.2

Re: [PATCH] staging: goldfish_audio.c: sparse warning of incorrect type

[Greg Kroah-Hartman]

Adding Alan Cox, as he pushed this driver upstream...

On Sun, Aug 31, 2014 at 09:46:22PM +0530, Sudip Mukherjee wrote:
> fixed sparse warning of incorrect type in argument 1 and incorrect type in argument 2
> it was directly dereferencing a __iomem pointer , which will work in x86 but will fail in other architectures.

Please wrap your changelog comments at 72 columns...

> Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
> ---
> 
> hi,
> can you please reveiew the patch and check if the approach is correct. If it is correct then I can send another patch to correct the other similar warnings present in this file. 

I don't think this is correct, because:

> 
>  drivers/staging/goldfish/goldfish_audio.c | 11 +++++++++--
>  1 file changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/staging/goldfish/goldfish_audio.c b/drivers/staging/goldfish/goldfish_audio.c
> index c89d0b8..f9a13e7 100644
> --- a/drivers/staging/goldfish/goldfish_audio.c
> +++ b/drivers/staging/goldfish/goldfish_audio.c
> @@ -118,6 +118,7 @@ static ssize_t goldfish_audio_read(struct file *fp, char __user *buf,
>  	struct goldfish_audio *data = fp->private_data;
>  	int length;
>  	int result = 0;
> +	void *buffer;
>  
>  	if (!data->read_supported)
>  		return -ENODEV;
> @@ -133,9 +134,15 @@ static ssize_t goldfish_audio_read(struct file *fp, char __user *buf,
>  						AUDIO_READ_BUFFER_AVAILABLE);
>  
>  		/* copy data to user space */
> -		if (copy_to_user(buf, data->read_buffer, length))
> +		buffer = kzalloc(length, GFP_KERNEL);
> +		if (buffer == NULL)
> +			return -ENOMEM;
> +		memcpy_fromio(buffer, data->read_buffer, length);
> +		if (copy_to_user((void __user *)buf, buffer, length)) {
> +			kfree(buffer);
>  			return -EFAULT;
> -
> +		}
> +		kfree(buffer);

The goldfish platform is an "emulator" so iomem is (or should be?) the
same as "normal" memory here.  So I think this is a false-positive.

But I could be totally wrong here, Alan?

thanks,

greg k-h

Re: [PATCH] staging: goldfish_audio.c: sparse warning of incorrect type

[Dan Carpenter]

On Sun, Aug 31, 2014 at 09:46:22PM +0530, Sudip Mukherjee wrote:
> @@ -133,9 +134,15 @@ static ssize_t goldfish_audio_read(struct file *fp, char __user *buf,
>  						AUDIO_READ_BUFFER_AVAILABLE);
>  
>  		/* copy data to user space */
> -		if (copy_to_user(buf, data->read_buffer, length))
> +		buffer = kzalloc(length, GFP_KERNEL);
> +		if (buffer == NULL)
> +			return -ENOMEM;
> +		memcpy_fromio(buffer, data->read_buffer, length);
> +		if (copy_to_user((void __user *)buf, buffer, length)) {
                                 ^^^^^^^^^^^^^^^
This cast shouldn't be needed.

regards,
dan carpenter

Re: [PATCH] staging: goldfish_audio.c: sparse warning of incorrect type

[Sudip Mukherjee]

On Mon, Sep 01, 2014 at 12:35:14PM +0300, Dan Carpenter wrote:
> On Sun, Aug 31, 2014 at 09:46:22PM +0530, Sudip Mukherjee wrote:
> > @@ -133,9 +134,15 @@ static ssize_t goldfish_audio_read(struct file *fp, char __user *buf,
> >  						AUDIO_READ_BUFFER_AVAILABLE);
> >  
> >  		/* copy data to user space */
> > -		if (copy_to_user(buf, data->read_buffer, length))
> > +		buffer = kzalloc(length, GFP_KERNEL);
> > +		if (buffer == NULL)
> > +			return -ENOMEM;
> > +		memcpy_fromio(buffer, data->read_buffer, length);
> > +		if (copy_to_user((void __user *)buf, buffer, length)) {
>                                  ^^^^^^^^^^^^^^^
> This cast shouldn't be needed.
> 
> regards,
> dan carpenter
> 

yes , its not needed . I just checked it again after removing the cast. Actually I gave it as sparse was showing warning for argument 1 also. I should have rechecked it after fixing the warning for argument2. Sorry for it.
Anyways, if Alan confirms that this patch is actually needed then I will resubmit a modified patch.

thanks
Sudip

Re: [PATCH] staging: goldfish_audio.c: sparse warning of incorrect type

[Alan Cox]

On Sun, 2014-08-31 at 13:02 -0700, Greg Kroah-Hartman wrote:
> Adding Alan Cox, as he pushed this driver upstream...

It's sort of a false positive. The existing code will work fine. 

Arguably all of this should be using the dma_ APIs.

Alan