[Cluster-devel] [PATCH] tunegfs2: Add some malloc error checking

[Steven Whitehouse <swhiteho@redhat.com>]

Hi,

Looks good,

Steve.

On Thu, 2011-08-04 at 10:49 +0100, Andrew Price wrote:
> Static analysis found a possible null pointer dereference due to a
> missing check in read_super. The malloc'd memory was also not being
> freed on error conditions. This patch adds a check for a null pointer
> and frees the allocated memory.
> 
> Signed-off-by: Andrew Price <anprice@redhat.com>
> ---
>  gfs2/tune/super.c |    7 +++++++
>  1 files changed, 7 insertions(+), 0 deletions(-)
> 
> diff --git a/gfs2/tune/super.c b/gfs2/tune/super.c
> index 65e8d5b..49d87d6 100644
> --- a/gfs2/tune/super.c
> +++ b/gfs2/tune/super.c
> @@ -94,14 +94,21 @@ int read_super(struct tunegfs2 *tfs)
>  	int n;
>         	tfs->sb_start = GFS2_SB_ADDR << GFS2_BASIC_BLOCK_SHIFT;
>  	block = malloc(sizeof(char) * GFS2_DEFAULT_BSIZE);
> +	if (!block) {
> +		perror("read_super: malloc");
> +		return EX_UNAVAILABLE;
> +	}
>  	n = pread(tfs->fd, block, GFS2_DEFAULT_BSIZE, tfs->sb_start);
>  	if (n < 0) {
>  		perror("read_super: pread");
> +		free(block);
>  		return EX_IOERR;
>  	}
>  	tfs->sb = block;
>  	if (be32_to_cpu(tfs->sb->sb_header.mh_magic) != GFS2_MAGIC) {
>  		fprintf(stderr, _("Not a GFS/GFS2 device\n"));
> +		tfs->sb = NULL;
> +		free(block);
>  		return EX_IOERR;
>  	}
>  	/* Ensure that table and proto are NULL terminated */