Received-SPF: pass (google.com: domain of linux-bluetooth-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210831115637.6713-1-kiran.k@intel.com> <20210831115637.6713-12-kiran.k@intel.com>
 <CABBYNZL3=z6CWj9EAiGnOkcaEmgs1Mq7kc5-fT4Mjk2HCUPsFA@mail.gmail.com> <135FFE98-549B-4C02-9AC3-AD7F18EC0564@holtmann.org>
In-Reply-To: <135FFE98-549B-4C02-9AC3-AD7F18EC0564@holtmann.org>
From:   Luiz Augusto von Dentz <luiz.dentz@gmail.com>
Date:   Thu, 2 Sep 2021 15:37:54 -0700
Message-ID: <CABBYNZJGF+VzeouaJXf7_T1smtgEqwyUimsxV8ajUZURMGBxfg@mail.gmail.com>
Subject: Re: [PATCH v13 12/12] Bluetooth: Allow usb to auto-suspend when SCO
 use non-HCI transport
To:     Marcel Holtmann <marcel@holtmann.org>
Cc:     Kiran K <kiran.k@intel.com>,
        "linux-bluetooth@vger.kernel.org" <linux-bluetooth@vger.kernel.org>,
        Chethan T N <chethan.tumkur.narayan@intel.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi Marcel,

On Thu, Sep 2, 2021 at 2:37 AM Marcel Holtmann <marcel@holtmann.org> wrote:
>
> Hi Luiz,
>
> >> Currently usb tranport is not allowed to suspend when SCO over
> >> HCI tranport is active.
> >>
> >> This patch shall enable the usb tranport to suspend when SCO
> >> link use non-HCI transport
> >>
> >> Signed-off-by: Chethan T N <chethan.tumkur.narayan@intel.com>
> >> Signed-off-by: Kiran K <kiran.k@intel.com>
> >> ---
> >>
> >> Notes:
> >>    changes in v13:
> >>    - suspend usb in HFP offload use case
> >>
> >> drivers/bluetooth/btintel.c       |  2 +-
> >> include/net/bluetooth/bluetooth.h |  4 ++++
> >> net/bluetooth/hci_event.c         | 20 +++++++++++---------
> >> net/bluetooth/sco.c               |  2 +-
> >> 4 files changed, 17 insertions(+), 11 deletions(-)
> >>
> >> diff --git a/drivers/bluetooth/btintel.c b/drivers/bluetooth/btintel.c
> >> index 6091b691ddc2..2d64e289cf6e 100644
> >> --- a/drivers/bluetooth/btintel.c
> >> +++ b/drivers/bluetooth/btintel.c
> >> @@ -2215,7 +2215,7 @@ static int btintel_get_codec_config_data(struct =
hci_dev *hdev,
> >> static int btintel_get_data_path_id(struct hci_dev *hdev, __u8 *data_p=
ath_id)
> >> {
> >>        /* Intel uses 1 as data path id for all the usecases */
> >> -       *data_path_id =3D 1;
> >> +       *data_path_id =3D BT_SCO_PCM_PATH;
> >>        return 0;
> >> }
> >>
> >> diff --git a/include/net/bluetooth/bluetooth.h b/include/net/bluetooth=
/bluetooth.h
> >> index c1fa90fb7502..9e2745863b33 100644
> >> --- a/include/net/bluetooth/bluetooth.h
> >> +++ b/include/net/bluetooth/bluetooth.h
> >> @@ -177,6 +177,10 @@ struct bt_codecs {
> >> #define CODING_FORMAT_TRANSPARENT      0x03
> >> #define CODING_FORMAT_MSBC             0x05
> >>
> >> +/* Audio data transport path used for SCO */
> >> +#define BT_SCO_HCI_PATH 0x00
> >> +#define BT_SCO_PCM_PATH 0x01
> >
> > If this is in fact vendor specific perhaps you should be declared in
> > btintel.h not here.
> >
> >> +
> >> __printf(1, 2)
> >> void bt_info(const char *fmt, ...);
> >> __printf(1, 2)
> >> diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
> >> index b48e24629fa4..7ff11cba89cf 100644
> >> --- a/net/bluetooth/hci_event.c
> >> +++ b/net/bluetooth/hci_event.c
> >> @@ -4516,15 +4516,17 @@ static void hci_sync_conn_complete_evt(struct =
hci_dev *hdev,
> >>
> >>        bt_dev_dbg(hdev, "SCO connected with air mode: %02x", ev->air_m=
ode);
> >>
> >> -       switch (ev->air_mode) {
> >> -       case 0x02:
> >> -               if (hdev->notify)
> >> -                       hdev->notify(hdev, HCI_NOTIFY_ENABLE_SCO_CVSD)=
;
> >> -               break;
> >> -       case 0x03:
> >> -               if (hdev->notify)
> >> -                       hdev->notify(hdev, HCI_NOTIFY_ENABLE_SCO_TRANS=
P);
> >> -               break;
> >> +       if (conn->codec.data_path =3D=3D BT_SCO_HCI_PATH) {
> >> +               switch (ev->air_mode) {
> >> +               case 0x02:
> >> +                       if (hdev->notify)
> >> +                               hdev->notify(hdev, HCI_NOTIFY_ENABLE_S=
CO_CVSD);
> >> +                       break;
> >> +               case 0x03:
> >> +                       if (hdev->notify)
> >> +                               hdev->notify(hdev, HCI_NOTIFY_ENABLE_S=
CO_TRANSP);
> >> +                       break;
> >> +               }
> >
> > Hmm I think we might need to notify the driver to enable PCM routing
> > so the likes of btusb can call
> > usb_disable_endpoint/usb_enable_endpoint for example since in theory
> > userspace may choose to switch from software to hardware offload and
> > vice-versa, note without calling usb_disable_endpoint there might not
> > be much power saving after all since the endpoint will remain active
> > or do we actually have a good explanation why it shall not be called
> > when using PCM routing? Note that usb_set_interface will call
> > usb_enable_interface that will then call usb_enable_endpoint so
> > perhaps we need to call usb_disable_interface, either way we can't
> > assume the platform will be restricted to only use one or the other.
>
> actually for the Intel hardware we shouldn=E2=80=99t do this at all. We s=
hould switch to vendor specific SCO over bulk endpoints and not claim the I=
SOC endpoints at all.

Yep, but I guess that requires switching to SCO over bulk then which
perhaps needs more changes, not sure if we should pursue that or go
with all the way with H4 mode like we did in Zephyr, anyway for the
purpose of offload I would be fine skipping the SCO over bulk since we
are already at v13 of these.

> Regards
>
> Marcel

>


--
Luiz Augusto von Dentz