PulseAudio Marge Bot pushed to branch master at PulseAudio / pulseaudio
Commits: d517a204 by Igor V. Kovalenko at 2021-08-25T15:00:28+00:00 bluetooth: Allow SCO socket packet size to be larger than mSBC frame With USB Alternate Setting 3 size of HCI payload is 72 bytes which is already larger than mSBC frame size. Largest known size of HCI payload is with USB Alternate Setting 5 (144 bytes), make it the default SCO socket MTU. Reserve additional space in bluetooth encoder buffer to cover this case. Since mSBC encoder and decoder will now work with larger packet sizes, drop assertions about MTU larger than mSBC frame size. Part-of: <https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/617> - - - - - 3 changed files: - src/modules/bluetooth/backend-native.c - src/modules/bluetooth/backend-ofono.c - src/modules/bluetooth/module-bluez5-device.c Changes: ===================================== src/modules/bluetooth/backend-native.c ===================================== @@ -329,8 +329,18 @@ static int sco_acquire_cb(pa_bluetooth_transport *t, bool optional, size_t *imtu if (sock < 0) goto fail; - if (imtu) *imtu = 60; - if (omtu) *omtu = 60; + /* The correct block size should take into account the SCO MTU from + * the Bluetooth adapter and (for adapters in the USB bus) the MxPS + * value from the Isoc USB endpoint in use by btusb and should be + * made available to userspace by the Bluetooth kernel subsystem. + * + * Set initial MTU to max known payload length of HCI packet + * in USB Alternate Setting 5 (144 bytes) + * See also pa_bluetooth_transport::last_read_size handling + * and comment about MTU size in bt_prepare_encoder_buffer() + */ + if (imtu) *imtu = 144; + if (omtu) *omtu = 144; if (t->device->autodetect_mtu) { struct sco_options sco_opt; @@ -347,11 +357,6 @@ static int sco_acquire_cb(pa_bluetooth_transport *t, bool optional, size_t *imtu } } - /* read/decode machinery only works if we get at most one MSBC encoded packet at a time - * when it is fixed to process stream of packets, lift this assertion */ - pa_assert(*imtu <= MSBC_PACKET_SIZE); - pa_assert(*omtu <= MSBC_PACKET_SIZE); - return sock; fail: ===================================== src/modules/bluetooth/backend-ofono.c ===================================== @@ -330,13 +330,15 @@ static int hf_audio_agent_transport_acquire(pa_bluetooth_transport *t, bool opti * value from the Isoc USB endpoint in use by btusb and should be * made available to userspace by the Bluetooth kernel subsystem. * - * Set initial MTU to max size which is reported to be working (60 bytes) - * See also pa_bluetooth_transport::last_read_size handling. + * Set initial MTU to max known payload length of HCI packet + * in USB Alternate Setting 5 (144 bytes) + * See also pa_bluetooth_transport::last_read_size handling + * and comment about MTU size in bt_prepare_encoder_buffer() */ if (imtu) - *imtu = 60; + *imtu = 144; if (omtu) - *omtu = 60; + *omtu = 144; err = socket_accept(card->fd); if (err < 0) { ===================================== src/modules/bluetooth/module-bluez5-device.c ===================================== @@ -260,25 +260,48 @@ static void connect_ports(struct userdata *u, void *new_data, pa_direction_t dir static bool bt_prepare_encoder_buffer(struct userdata *u) { - size_t encoded_size, reserved_size; + size_t encoded_size, reserved_size, encoded_frames; pa_assert(u); pa_assert(u->bt_codec); /* If socket write MTU is less than encoded frame size, there could be * up to one write MTU of data left in encoder buffer from previous round. * - * Reserve space for 2 encoded frames to cover that. + * Reserve space for at least 2 encoded frames to cover that. * * Note for A2DP codecs it is expected that size of encoded frame is less * than write link MTU. Therefore each encoded frame is sent out completely * and there is no used space in encoder buffer before next encoder call. + * + * For SCO socket all writes will be of MTU size to match payload length + * of HCI packet. Depending on selected USB Alternate Setting the payload + * length of HCI packet may exceed encoded frame size. For mSBC frame size + * is 60 bytes, payload length of HCI packet in USB Alts 3 is 72 byte, + * in USB Alts 5 it is 144 bytes. + * + * Reserve space for up to 1 + MTU / (encoded frame size) encoded frames + * to cover that. + * + * Note for current linux kernel (up to 5.13.x at least) there is no way to + * reliably detect socket MTU size. For now we just set SCO socket MTU to be + * large enough to cover all known sizes (largest is USB ALts 5 with 144 bytes) + * and adjust SCO write size to be equal to last SCO read size. This makes + * write size less or equal to MTU size. Reserving the same number of encoded + * frames to cover full MTU is still enough. + * See also https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/merge_requests/254#note_779802 */ + if (u->bt_codec->get_encoded_block_size) encoded_size = u->bt_codec->get_encoded_block_size(u->encoder_info, u->write_block_size); else encoded_size = u->write_block_size; - reserved_size = 2 * encoded_size; + encoded_frames = u->write_link_mtu / u->write_block_size + 1; + + if (encoded_frames < 2) + encoded_frames = 2; + + reserved_size = encoded_frames * encoded_size; if (u->encoder_buffer_size < reserved_size) { u->encoder_buffer = pa_xrealloc(u->encoder_buffer, reserved_size); View it on GitLab: https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/commit/d517a204de70432d6783d103ab650d62d8c1b051 -- View it on GitLab: https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/commit/d517a204de70432d6783d103ab650d62d8c1b051 You're receiving this email because of your account on gitlab.freedesktop.org.
