Re: [PATCH] dma-buf: Fix memory leak in dma_buf_set_name

[Ezequiel Garcia]

Hi Bharath,

Thanks for taking the time to try to fix this report.

However, this doesn't look right.

On Fri, 2019-08-16 at 23:30 +0530, Bharath Vedartham wrote:
> This patch fixes a memory leak bug reported by syzbot. Link to the
> bug is given at [1].
> 
> A local variable name is used to hold the copied user buffer string
> using strndup_user. strndup_user allocates memory using
> kmalloc_track_caller in memdup_user. This kmalloc allocation needs to be
> followed by a kfree.
> 
> This patch has been tested by a compile test.
> 
> [1] https://syzkaller.appspot.com/bug?id=ce692a3aa13e00e335e090be7846c6eb60ddff7a
> 
> Reported-by: syzbot+b2098bc44728a4efb3e9@xxxxxxxxxxxxxxxxxxxxxxxxx
> Signed-off-by: Bharath Vedartham <linux.bhar@xxxxxxxxx>
> ---
>  drivers/dma-buf/dma-buf.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c
> index f45bfb2..9798f6d 100644
> --- a/drivers/dma-buf/dma-buf.c
> +++ b/drivers/dma-buf/dma-buf.c
> @@ -342,6 +342,7 @@ static long dma_buf_set_name(struct dma_buf *dmabuf, const char __user *buf)
>  	}
>  	kfree(dmabuf->name);
>  	dmabuf->name = name;
> +	kfree(name);
>  

Just by looking at this, you can deduce something is not right.
You are assigning "name" to dmabuf->name, but then releasing "name"!

So now, dmabuf->name has free memory, which will lead to
user-after-free issues.

Note also, that this function doesn't look leaky since the previous
"name" is freed, before setting a new one.

Maybe the syzbot report is some kind of false positive?

Also, I _strongly_ suggest that in the future you don't compile-test
only these kind of not trivial fixes. Since you are touching a crucial
part of the kernel here, you should really be testing properly.

Specially since syzbot produces a reproducer.

Consider compile test as something you do when your changes are
only cosmetic, and you are completely and absolutely sure things
will be OK.

Thanks.
Ezequiel