Re: [PATCH v2 1/2] Bluetooth: Introduce HCI_CONN_FLAG_DEVICE_PRIVACY device flag

[Luiz Augusto von Dentz <luiz.dentz@gmail.com>]

Hi Marcel,

On Thu, Nov 18, 2021 at 9:19 AM Marcel Holtmann <marcel@holtmann.org> wrote:
>
> Hi Luiz,
>
> >>> This introduces HCI_CONN_FLAG_DEVICE_PRIVACY which can be used by
> >>> userspace to indicate to the controller to use Device Privacy Mode to a
> >>> specific device.
> >>>
> >>> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> >>> ---
> >>> v2: Fix marking Device Privacy Flag even when adapter is not capable of
> >>> handling Set Privacy Mode.
> >>>
> >>> include/net/bluetooth/hci_core.h |  4 ++++
> >>> net/bluetooth/mgmt.c             | 24 ++++++++++++++++++++----
> >>> 2 files changed, 24 insertions(+), 4 deletions(-)
> >>>
> >>> diff --git a/include/net/bluetooth/hci_core.h b/include/net/bluetooth/hci_core.h
> >>> index b5f061882c10..07d2d099dc2a 100644
> >>> --- a/include/net/bluetooth/hci_core.h
> >>> +++ b/include/net/bluetooth/hci_core.h
> >>> @@ -160,6 +160,7 @@ struct bdaddr_list_with_flags {
> >>>
> >>> enum hci_conn_flags {
> >>>      HCI_CONN_FLAG_REMOTE_WAKEUP,
> >>> +     HCI_CONN_FLAG_DEVICE_PRIVACY,
> >>>      HCI_CONN_FLAG_MAX
> >>> };
> >>>
> >>> @@ -1468,6 +1469,9 @@ void hci_conn_del_sysfs(struct hci_conn *conn);
> >>> #define use_ll_privacy(dev) (ll_privacy_capable(dev) && \
> >>>                           hci_dev_test_flag(dev, HCI_ENABLE_LL_PRIVACY))
> >>>
> >>> +#define privacy_mode_capable(dev) (use_ll_privacy(dev) && \
> >>> +                                (hdev->commands[39] & 0x04))
> >>> +
> >>> /* Use enhanced synchronous connection if command is supported */
> >>> #define enhanced_sco_capable(dev) ((dev)->commands[29] & 0x08)
> >>>
> >>> diff --git a/net/bluetooth/mgmt.c b/net/bluetooth/mgmt.c
> >>> index 06384d761928..8a8376d32be3 100644
> >>> --- a/net/bluetooth/mgmt.c
> >>> +++ b/net/bluetooth/mgmt.c
> >>> @@ -4350,7 +4350,16 @@ static int set_exp_feature(struct sock *sk, struct hci_dev *hdev,
> >>>                             MGMT_STATUS_NOT_SUPPORTED);
> >>> }
> >>>
> >>> -#define SUPPORTED_DEVICE_FLAGS() ((1U << HCI_CONN_FLAG_MAX) - 1)
> >>> +static u32 supported_device_flags(struct hci_dev *hdev)
> >>> +{
> >>> +     u32 flags = BIT(HCI_CONN_FLAG_MAX) - 1;
> >>> +
> >>> +     /* Check if Privacy Mode can be set */
> >>> +     if (!privacy_mode_capable(hdev))
> >>> +             flags &= ~BIT(HCI_CONN_FLAG_DEVICE_PRIVACY);
> >>> +
> >>> +     return flags;
> >>> +}
> >>
> >> I am lost on what we are doing, I know that SUPPORTED_DEVICE_FLAGS was introduced by 4c54bf2b093bb from Abhishek, but I fail to reason now why it is correct.
> >
> > But we don't set the HCI_CONN_FLAG_DEVICE_PRIVACY is hdev is not
> > capable of setting it, anyway the general idea of the supported_flags
> > is to indicate to userspace what flags the kernel is capable of
> > settings, so yeah I would expected it to be capable of setting every
> > flag except for those the controller don't have proper support for.
> > Maybe you got confused by the logic of first enabling everything and
> > then toggle back the bits that are not supported by the hdev.
>
> and I think that is wrong. We should only set the bits that the hardware and/or kernel supports. I have no idea on how I missed this in the review back then. I am trying to figure it out, but I am failing to follow this logic.

There is only 1 flag currently though, HCI_CONN_FLAG_REMOTE_WAKEUP, so
I wonder if that really requires a check? Or perhaps you want to
change the way we supported_device_flags works to:

index fe52b7eefb56..3ad09ce6eaeb 100644
--- a/net/bluetooth/mgmt.c
+++ b/net/bluetooth/mgmt.c
@@ -4351,11 +4351,15 @@ static int set_exp_feature(struct sock *sk,
struct hci_dev *hdev,

 static u32 supported_device_flags(struct hci_dev *hdev)
 {
-       u32 flags = BIT(HCI_CONN_FLAG_MAX) - 1;
+       u32 flags = 0;
+
+       /* Check if adapter can wakeup the system */
+       if (hdev->wakeup && hdev->wakeup(hdev))
+               flags |= HCI_CONN_FLAG_REMOTE_WAKEUP;

        /* Check if Privacy Mode can be set */
-       if (!privacy_mode_capable(hdev))
-               flags &= ~BIT(HCI_CONN_FLAG_DEVICE_PRIVACY);
+       if (privacy_mode_capable(hdev))
+               flags |= HCI_CONN_FLAG_DEVICE_PRIVACY;

        return flags;
 }

We will need to adjust some of the tests though to reflect this change
in behavior:

et Device Flags - Success                           Failed       0.016 seconds
Set Device Flags - Success                           Not Run