Re: [PATCH v2 1/3] eir: parse data types for LE OOB pairing

[Mike Brudevold <puffy.taco@gmail.com>]

Hi Luiz,

On Mon, Jun 13, 2022 at 4:55 PM Luiz Augusto von Dentz
<luiz.dentz@gmail.com> wrote:
>
> Hi Mike,
>
> On Mon, Jun 13, 2022 at 2:42 PM Mike Brudevold <puffy.taco@gmail.com> wrote:
> >
> > Hi Luiz,
> >
> > > It might be better to handle this via bt_ad instance instead of
> > > eir_data, in fact the plan was always to switch to bt_ad but it seems
> > > we forgot about it at some point.
> >
> > Are you thinking something like below (doesn't fully compile,
> > name2utf8 is static in eir so I did nothing about that yet)?
> > Basically where the ad code parses the EIR data, but the neard plugin
> > still manages what to do with the data?  The alternative would be
> > where device.c became smarter to consume the ad struct itself and the
> > neard plugin becomes a simple conduit for the ad data.
>
> The later is probably better alternative, like I said I wrote bt_ad to
> replace eir handling completely so we could also write proper unit
> testing for it, Im fine if you want to take the time to change the
> daemon core itself but at least introduce support for the types you
> will be using in the plugin and then make use of them.
>
> > index 77a4630da..3d4064515 100644
> > --- a/plugins/neard.c
> > +++ b/plugins/neard.c
> > @@ -31,6 +31,7 @@
> >  #include "src/agent.h"
> >  #include "src/btd.h"
> >  #include "src/shared/util.h"
> > +#include "src/shared/ad.h"
> >
> >  #define NEARD_NAME "org.neard"
> >  #define NEARD_PATH "/org/neard"
> > @@ -336,6 +337,52 @@ static int check_device(struct btd_device *device)
> >         return 0;
> >  }
> >
> > +static void process_oob_adv(void *data, void *user_data)
> > +{
> > +       struct bt_ad_data *ad_data = data;
> > +       struct oob_params *remote = user_data;
> > +       uint8_t name_len;
> > +
> > +       switch (ad_data->type) {
> > +       case EIR_NAME_SHORT:
> > +       case EIR_NAME_COMPLETE:
> > +               name_len = ad_data->len;
> > +
> > +               /* Some vendors put a NUL byte terminator into
> > +                       * the name */
> > +               while (name_len > 0 && ad_data->data[name_len - 1] == '\0')
> > +                       name_len--;
> > +
> > +               g_free(remote->name);
> > +
> > +               remote->name = name2utf8(ad_data->data, name_len);
> > +               break;
> > +
> > +       case BT_AD_LE_DEVICE_ADDRESS:
> > +               if (ad_data->len < sizeof(bdaddr_t) + 1)
> > +                       break;
> > +
> > +               memcpy(&remote->address, ad_data->data, sizeof(bdaddr_t));
> > +               remote->address_type = ad_data->data[sizeof(bdaddr_t)] & 0x1 ?
> > +                               BDADDR_LE_RANDOM : BDADDR_LE_PUBLIC;
> > +               break;
> > +
> > +       case EIR_LE_SC_CONF:
> > +               if (ad_data->len < 16)
> > +                       break;
> > +               free(remote->hash256);
> > +               remote->hash256 = util_memdup(ad_data->data, 16);
> > +               break;
> > +
> > +       case EIR_LE_SC_RAND:
> > +               if (ad_data->len < 16)
> > +                       break;
> > +               free(remote->randomizer256);
> > +               remote->randomizer256 = util_memdup(ad_data->data, 16);
> > +               break;
> > +       }
> > +}
>
> Do we need to duplicate this information? I'd consider just using the
> bt_ad instance to parse and store them, well perhaps we want to
> introduce something like bt_ad_foreach_type so you can locate the data
> by type?

That's partly what I was checking on.  The ad code doesn't have much
functionality right now to be able to parse the meaning of the data,
beyond storing them in TLV format (bt_ad_data).  Which is the opposite
to how the data is given to ad if you're creating an advertisement
(e.g., service UUIDs are stored in bt_uuid_t format inside the service
queue when using bt_ad_add_service_uuid, not in the data queue).  This
means the ad code likely has to get smarter, but I wanted to make sure
I wasn't missing something that should have been obvious.  If the ad
code can present the data back in a usable format, then it wouldn't
really have to be duplicated.  Otherwise this code would have been an
easy way to not use the eir code while the ad code gets developed.
With some concern still that there's a type_reject_list related to the
data ad queue, but that can be completely bypassed when using
bt_ad_new_with_data - so this method is doing something that seems
unintended.

>
> >  static int process_eir(uint8_t *eir, size_t size, struct oob_params *remote)
> >  {
> >         struct eir_data eir_data;
> > @@ -370,32 +417,17 @@ static int process_eir(uint8_t *eir, size_t
> > size, struct oob_params *remote)
> >
> >  static void process_eir_le(uint8_t *eir, size_t size, struct
> > oob_params *remote)
> >  {
> > -       struct eir_data eir_data;
> > +       struct bt_ad *ad;
> >
> >         DBG("size %zu", size);
> >
> > -       memset(&eir_data, 0, sizeof(eir_data));
> > -
> > -       eir_parse(&eir_data, eir, size);
> > -
> > -       bacpy(&remote->address, &eir_data.addr);
> > -       remote->address_type = eir_data.addr_type;
> > -
> > -       remote->class = eir_data.class;
> > -
> > -       remote->name = eir_data.name;
> > -       eir_data.name = NULL;
> > -
> > -       remote->services = eir_data.services;
> > -       eir_data.services = NULL;
> > -
> > -       remote->hash256 = eir_data.hash256;
> > -       eir_data.hash256 = NULL;
> > +       ad = bt_ad_new_with_data(size, eir);
> >
> > -       remote->randomizer256 = eir_data.randomizer256;
> > -       eir_data.randomizer256 = NULL;
> > +       if (ad) {
> > +               bt_ad_foreach_data(ad, process_oob_adv, remote);
> >
> > -       eir_data_free(&eir_data);
> > +               bt_ad_unref(ad);
> > +       }
> >  }
>
>
>
> --
> Luiz Augusto von Dentz