possible memory leak in commands/nand.c?

[Robert P. J. Day]

On Mon, 21 Dec 2009, Sascha Hauer wrote:

... snip ...

> Yes, indeed, that's a memory hole here. The following should fix
> this. Thanks for noting.
>
> Sascha
>
>
> >From 4e4b03cd61808383a98cb1d10a47025e1909e0bd Mon Sep 17 00:00:00 2001
> From: Sascha Hauer <s.hauer at pengutronix.de>
> Date: Mon, 21 Dec 2009 09:41:52 +0100
> Subject: [PATCH] commands/nand.c: Fix memory hole
>
> Signed-off-by: Sascha Hauer <s.hauer at pengutronix.de>
> ---
>  commands/nand.c |   22 +++++++++++++++++-----
>  1 files changed, 17 insertions(+), 5 deletions(-)
>
> diff --git a/commands/nand.c b/commands/nand.c
> index cbf1058..55b89af 100644
> --- a/commands/nand.c
> +++ b/commands/nand.c
> @@ -224,31 +224,37 @@ static struct file_operations nand_bb_ops = {
>  int dev_add_bb_dev(char *path, const char *name)
>  {
>  	struct nand_bb *bb;
> -	int ret;
> +	int ret = -ENOMEM;
>  	struct stat s;
>
>  	bb = xzalloc(sizeof(*bb));
>  	bb->devname = asprintf("/dev/%s", basename(path));
> +	if (!bb->devname)
> +		goto out1;
> +
>  	if (name)
>  		bb->cdev.name = strdup(name);
>  	else
>  		bb->cdev.name = asprintf("%s.bb", basename(path));
>
> +	if (!bb->cdev.name)
> +		goto out2;
> +
>  	ret = stat(bb->devname, &s);
>  	if (ret)
> -		goto free_out;
> +		goto out3;
>
>  	bb->raw_size = s.st_size;
>
>  	bb->fd = open(bb->devname, O_RDWR);
>  	if (bb->fd < 0) {
>  		ret = -ENODEV;
> -		goto free_out;
> +		goto out3;
>  	}
>
>  	ret = ioctl(bb->fd, MEMGETINFO, &bb->info);
>  	if (ret)
> -		goto free_out;
> +		goto out4;
>
>  	nand_bb_calc_size(bb);
>  	bb->cdev.ops = &nand_bb_ops;
> @@ -258,7 +264,13 @@ int dev_add_bb_dev(char *path, const char *name)
>
>  	return 0;
>
> -free_out:
> +out4:
> +	close(bb->fd);
> +out3:
> +	free(bb->cdev.name);
> +out2:
> +	free(bb->devname);
> +out1:
>  	free(bb);
>  	return ret;
>  }

  i'm not sure this required distinguishing between every one of those
cases since the initial space was allocated with xzalloc(),
guaranteeing it would be zero-filled, and freeing a NULL pointer is
supposed to be a no-op.

  so it would have been simpler to just

  free(bb->devname);		# might be NULL, no problem
  free(bb->cdev.name);		# same here
  free(bb);

but, yes, the above will work.

  there's another memory leak i've found, i'll submit a patch for it,
for no other reason than i feel like getting a few patches with my
name on it into the barebox git log. :-)

rday
--


========================================================================
Robert P. J. Day                               Waterloo, Ontario, CANADA

            Linux Consulting, Training and Kernel Pedantry.

Web page:                                          http://crashcourse.ca
Twitter:                                       http://twitter.com/rpjday
========================================================================