Hi Marcel,
On Sat, Jan 4, 2020 at 1:56 AM Marcel Holtmann <marcel@xxxxxxxxxxxx> wrote:
>
> Hi Luiz,
>
> > This adds BT_PHYS socket option which can be used to read the PHYs in
> > use by the underline connection.
> >
> > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@xxxxxxxxx>
> > ---
> > include/net/bluetooth/bluetooth.h | 17 ++++++++
> > include/net/bluetooth/hci_core.h | 2 +
> > net/bluetooth/hci_conn.c | 64 +++++++++++++++++++++++++++++++
> > net/bluetooth/l2cap_sock.c | 13 +++++++
> > net/bluetooth/sco.c | 13 +++++++
> > 5 files changed, 109 insertions(+)
> >
> > diff --git a/include/net/bluetooth/bluetooth.h b/include/net/bluetooth/bluetooth.h
> > index e42bb8e03c09..69c0e7eb26d9 100644
> > --- a/include/net/bluetooth/bluetooth.h
> > +++ b/include/net/bluetooth/bluetooth.h
> > @@ -121,6 +121,23 @@ struct bt_voice {
> >
> > #define BT_SNDMTU 12
> > #define BT_RCVMTU 13
> > +#define BT_PHYS 14
> > +
> > +#define BT_PHY_BR_1M_1SLOT 0x00000001
> > +#define BT_PHY_BR_1M_3SLOT 0x00000002
> > +#define BT_PHY_BR_1M_5SLOT 0x00000004
> > +#define BT_PHY_EDR_2M_1SLOT 0x00000008
> > +#define BT_PHY_EDR_2M_3SLOT 0x00000010
> > +#define BT_PHY_EDR_2M_5SLOT 0x00000020
> > +#define BT_PHY_EDR_3M_1SLOT 0x00000040
> > +#define BT_PHY_EDR_3M_3SLOT 0x00000080
> > +#define BT_PHY_EDR_3M_5SLOT 0x00000100
> > +#define BT_PHY_LE_1M_TX 0x00000200
> > +#define BT_PHY_LE_1M_RX 0x00000400
> > +#define BT_PHY_LE_2M_TX 0x00000800
> > +#define BT_PHY_LE_2M_RX 0x00001000
> > +#define BT_PHY_LE_CODED_TX 0x00002000
> > +#define BT_PHY_LE_CODED_RX 0x00004000
> >
> > __printf(1, 2)
> > void bt_info(const char *fmt, ...);
> > diff --git a/include/net/bluetooth/hci_core.h b/include/net/bluetooth/hci_core.h
> > index faebe3859931..03cf3f0f22b9 100644
> > --- a/include/net/bluetooth/hci_core.h
> > +++ b/include/net/bluetooth/hci_core.h
> > @@ -1467,6 +1467,8 @@ void *hci_sent_cmd_data(struct hci_dev *hdev, __u16 opcode);
> > struct sk_buff *hci_cmd_sync(struct hci_dev *hdev, u16 opcode, u32 plen,
> > const void *param, u32 timeout);
> >
> > +u32 hci_conn_get_phys(struct hci_conn *conn);
> > +
> > /* ----- HCI Sockets ----- */
> > void hci_send_to_sock(struct hci_dev *hdev, struct sk_buff *skb);
> > void hci_send_to_channel(unsigned short channel, struct sk_buff *skb,
> > diff --git a/net/bluetooth/hci_conn.c b/net/bluetooth/hci_conn.c
> > index 87691404d0c6..386e6b0bd2ab 100644
> > --- a/net/bluetooth/hci_conn.c
> > +++ b/net/bluetooth/hci_conn.c
> > @@ -1725,3 +1725,67 @@ struct hci_chan *hci_chan_lookup_handle(struct hci_dev *hdev, __u16 handle)
> >
> > return hchan;
> > }
> > +
> > +u32 hci_conn_get_phys(struct hci_conn *conn)
> > +{
> > + u32 phys = 0;
> > +
> > + hci_dev_lock(conn->hdev);
> > +
> > + switch (conn->type) {
> > + case ACL_LINK:
> > + case SCO_LINK:
> > + phys |= BT_PHY_BR_1M_1SLOT;
> > +
> > + if (conn->pkt_type & (HCI_DM3 | HCI_DH3))
> > + phys |= BT_PHY_BR_1M_3SLOT;
> > +
> > + if (conn->pkt_type & (HCI_DM5 | HCI_DH5))
> > + phys |= BT_PHY_BR_1M_5SLOT;
> > +
> > + if (!(conn->pkt_type & HCI_2DH1))
> > + phys |= BT_PHY_EDR_2M_1SLOT;
> > +
> > + if (!(conn->pkt_type & HCI_2DH3))
> > + phys |= BT_PHY_EDR_2M_3SLOT;
> > +
> > + if (!(conn->pkt_type & HCI_2DH5))
> > + phys |= BT_PHY_EDR_2M_5SLOT;
> > +
> > + if (!(conn->pkt_type & HCI_3DH1))
> > + phys |= BT_PHY_EDR_3M_1SLOT;
> > +
> > + if (!(conn->pkt_type & HCI_3DH3))
> > + phys |= BT_PHY_EDR_3M_3SLOT;
> > +
> > + if (!(conn->pkt_type & HCI_3DH5))
> > + phys |= BT_PHY_EDR_3M_5SLOT;
> > +
> > + break;
>
> Actually ACL packet types and eSCO packet types are different. You need to split this into packet helpers for L2CAP BR/EDR and LE and SCO/eSCO packet types.
Right the SCO and ESCO actually use EV packet types, though the
question is then should we map them to new bit-fields on BT_PHY
namespace or just use the same defines (RATE+SLOT) since the socket
family should be enough to indicate if they are EV rather than DH
packets? @Pali do you need the exact packet type for sco/esco, afaik
there is not so much use of them in that case because the codecs are
normally fixed rate over HFP.
> I hope we actually do this all correctly to track the packets types. Especially since the host just sets the allowed packet types and the controller can choose whatever type they want.
At least we do seem to store them properly for ACL.
> Regards
>
> Marcel
>
--
Luiz Augusto von Dentz
