Re: [PATCH] target: add error handling for match_int

From: James Bottomley
Date: Tue Jun 12 2018 - 17:21:31 EST


On Tue, 2018-06-12 at 12:52 +0800, Zhouyang Jia wrote:
> When match_int fails, the lack of error-handling code may
> cause unexpected results.
>
> This patch adds error-handling code after calling match_int.
>
> Signed-off-by: Zhouyang Jia <jiazhouyang09@xxxxxxxxx>
> ---
> Âdrivers/target/target_core_rd.c | 6 ++++--
> Â1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/target/target_core_rd.c
> b/drivers/target/target_core_rd.c
> index a6e8106..7bc89ff 100644
> --- a/drivers/target/target_core_rd.c
> +++ b/drivers/target/target_core_rd.c
> @@ -573,14 +573,16 @@ static ssize_t
> rd_set_configfs_dev_params(struct se_device *dev,
> Â token = match_token(ptr, tokens, args);
> Â switch (token) {
> Â case Opt_rd_pages:
> - match_int(args, &arg);
> + if (match_int(args, &arg))
> + return -EINVAL;

The first observation is that this would leak the kmalloc'd orig
variable, but the second is that I don't think terminating parsing is
the right thing to do even if match_int() returns an error: just
ignoring this option and proceed to the next seems to be the best
course because that's what we do with unrecognised options (the
default: case).

James