Applied
On Wed, 2020-03-25 at 19:12 +0000, Gix, Brian wrote:
> Hi Prathyusha,
>
> Inga and I have been talking, and comparing behavior to spec, and we believe that the *only* way that invalid
> data can make its way into the system is to hand-edit the node.json file, which is not something we want to
> encourage. We suggest moving validation to where the node.json is read....
> mesh/config-json.c in parse_features()
>
> Basically, we only care if
> "mode":"enabled",
>
> in which case, we range check interval (10-320) and count (1-8).
>
> Any values outside that range, we "Fail to Parse" the node, and that node will not be loaded.... So edit the
> node.json by hand *only* with legal arguments.
>
> Beyond that, if relay is "disabled" or "unsupported", the interval and count are don't cares, and there is no
> obligation for the 2nd parameter octet of RELAY_STATUS to be zero'd. As for the incoming RELAY_SET, there
> are
> no out of range or disallowed values for count or interval.
>
>
> On Wed, 2020-03-25 at 00:57 +0530, Prathyusha Nelluri wrote:
> > From: Prathyusha N <prathyusha.n@xxxxxxxxxxx>
> >
> > Added limit checking condition for count and interval
> > before processing for count and interval steps.
> > ---
> > mesh/cfgmod-server.c | 6 +++++-
> > 1 file changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/mesh/cfgmod-server.c b/mesh/cfgmod-server.c
> > index 7111411c7..151cab154 100644
> > --- a/mesh/cfgmod-server.c
> > +++ b/mesh/cfgmod-server.c
> > @@ -881,7 +881,11 @@ static bool cfg_srv_pkt(uint16_t src, uint32_t dst, uint16_t unicast,
> > n = mesh_model_opcode_set(OP_CONFIG_RELAY_STATUS, msg);
> >
> > msg[n++] = node_relay_mode_get(node, &count, &interval);
> > - msg[n++] = (count - 1) + ((interval/10 - 1) << 3);
> > +
> > + if (count > 0 && interval >= 10)
> > + msg[n++] = (count - 1) + ((interval/10 - 1) << 3);
> > + else
> > + msg[n++] = 0;
> >
> > l_debug("Get/Set Relay Config (%d)", msg[n-1]);
> > break;
