On Tue, 2013-09-03 at 21:18 -0300, Jo?o Paulo Rechi Vita wrote:
> On Sun, Aug 18, 2013 at 6:37 AM, Tanu Kaskinen
> <tanu.kaskinen at linux.intel.com> wrote:
> > On Tue, 2013-08-13 at 01:54 -0300, jprvita at gmail.com wrote:
> >> static DBusMessage *endpoint_select_configuration(DBusConnection *conn, DBusMessage *m, void *userdata) {
> >> + pa_bluetooth_discovery *y = userdata;
> >> + a2dp_sbc_t *cap, config;
> >> + uint8_t *pconf = (uint8_t *) &config;
> >> + int i, size;
> >> DBusMessage *r;
> >> + DBusError err;
> >>
> >> - pa_assert_se(r = dbus_message_new_error(m, BLUEZ_MEDIA_ENDPOINT_INTERFACE ".Error.NotImplemented",
> >> - "Method not implemented"));
> >> + static const struct {
> >> + uint32_t rate;
> >> + uint8_t cap;
> >> + } freq_table[] = {
> >> + { 16000U, SBC_SAMPLING_FREQ_16000 },
> >> + { 32000U, SBC_SAMPLING_FREQ_32000 },
> >> + { 44100U, SBC_SAMPLING_FREQ_44100 },
> >> + { 48000U, SBC_SAMPLING_FREQ_48000 }
> >> + };
> >> +
> >> + dbus_error_init(&err);
> >> +
> >> + if (!dbus_message_get_args(m, &err, DBUS_TYPE_ARRAY, DBUS_TYPE_BYTE, &cap, &size, DBUS_TYPE_INVALID)) {
> >
> > Passing &cap may cause alignment issues when accessing cap later. A byte
> > array pointer should be passed instead, and the memory should be copied
> > to an a2dp_sbc_t variable.
> >
>
> In my understanding this is the same case of your previous comment in
> enpoint_set_configuration(): a2dp_sbc_t is 1-byte aligned (packed and
> has only uint8_t fields), so there shouldn't be any problem. Or am I
> missing something?
No, you're not missing anything. My mistake.
> >> + pa_log_error("Endpoint SelectConfiguration(): %s", err.message);
> >> + dbus_error_free(&err);
> >> + goto fail;
> >> + }
> >> +
> >> + if (size != sizeof(config)) {
> >> + pa_log_error("Capabilities array has invalid size");
> >> + goto fail;
> >> + }
> >>
> >> + pa_zero(config);
> >> +
> >> + /* Find the lowest freq that is at least as high as the requested sampling rate */
> >> + for (i = 0; (unsigned) i < PA_ELEMENTSOF(freq_table); i++)
> >> + if (freq_table[i].rate >= y->core->default_sample_spec.rate && (cap->frequency & freq_table[i].cap)) {
> >> + config.frequency = freq_table[i].cap;
> >> + break;
> >> + }
> >> +
> >> + if ((unsigned) i == PA_ELEMENTSOF(freq_table)) {
> >> + for (--i; i >= 0; i--) {
> >> + if (cap->frequency & freq_table[i].cap) {
> >> + config.frequency = freq_table[i].cap;
> >> + break;
> >> + }
> >> + }
> >> +
> >> + if (i < 0) {
> >> + pa_log_error("Not suitable sample rate");
> >> + goto fail;
> >> + }
> >> + }
> >> +
> >> + pa_assert((unsigned) i < PA_ELEMENTSOF(freq_table));
> >> +
> >> + if (y->core->default_sample_spec.channels <= 1) {
> >> + if (cap->channel_mode & SBC_CHANNEL_MODE_MONO)
> >> + config.channel_mode = SBC_CHANNEL_MODE_MONO;
> >> + }
> >> +
> >> + if (y->core->default_sample_spec.channels >= 2) {
> >> + if (cap->channel_mode & SBC_CHANNEL_MODE_JOINT_STEREO)
> >> + config.channel_mode = SBC_CHANNEL_MODE_JOINT_STEREO;
> >> + else if (cap->channel_mode & SBC_CHANNEL_MODE_STEREO)
> >> + config.channel_mode = SBC_CHANNEL_MODE_STEREO;
> >> + else if (cap->channel_mode & SBC_CHANNEL_MODE_DUAL_CHANNEL)
> >> + config.channel_mode = SBC_CHANNEL_MODE_DUAL_CHANNEL;
> >> + else if (cap->channel_mode & SBC_CHANNEL_MODE_MONO) {
> >> + config.channel_mode = SBC_CHANNEL_MODE_MONO;
> >> + } else {
> >> + pa_log_error("No supported channel modes");
> >> + goto fail;
> >> + }
> >> + }
> >
> > You didn't address my earlier complaint: "If
> > default_sample_spec.channels is 1, but cap doesn't contain
> > SBC_CHANNEL_MODE_MONO, then config.channel_mode is left uninitialized
> > (well, it's initialized to zero in the beginning). I believe this is not
> > what you intended."
> >
>
> Yes, I think it is better to return an error to the caller here.
There's no need to return an error, we should allow any of the
two-channel modes even if default_sample_spec.channels is 1.
> > I also said that if default_sample_spec is 2 and cap only contains
> > SBC_CHANNEL_MODE_MONO, then the negotiation fails. It seems that that
> > complaint was bogus, though, so you're right to ignore that bit.
> >
>
> Sorry, I didn't understand the second part of your comment. If
> default_sample_spec.channels == 2 and cap->channel_mode has only
> SBC_CHANNEL_MODE_MONO, then config.channel_mode will receive
> SBC_CHANNEL_MODE_MONO.
Yes, this is what I meant by the complaint being bogus, so your code is
fine on this part.
--
Tanu
