Re: [PATCH BlueZ 4/4] input: Validate SDP HIDDescriptorList subattributes

[Johan Hedberg <johan.hedberg@gmail.com>]

Hi Lizardo,

On Sun, Jan 06, 2013, Anderson Lizardo wrote:
> It should not be assumed that remote SDP attributes are in a compliant
> format. This fixes a couple of invalid pointer access on invalid data.
> ---
>  profiles/input/device.c |   60 ++++++++++++++++++++++++++++++++++++-----------
>  1 file changed, 46 insertions(+), 14 deletions(-)

I've applied the first three patches, but this one needs a bit of fixing
up:

> +	if (d->dtd < SDP_SEQ8 || d->dtd > SDP_SEQ32)
> +		goto invalid_desc;

Please always be explicit on what values you're checking for instead of
assuming that the reader of the code knows what's contained within some
range. In this case there's already a convenient SDP_IS_SEQ() macro you
could use.

> +	if (d->dtd < SDP_SEQ8 || d->dtd > SDP_SEQ32)
> +		goto invalid_desc;

Same here.

> +	if (!d || d->dtd < SDP_TEXT_STR8 || d->dtd > SDP_TEXT_STR32)
> +		goto invalid_desc;

I suppose the best way to handle this one is to add a SDP_IS_STR() macro
(in a separate patch) to lib/sdp.h and then use it in this patch.

Johan