Re: [PATCH v4 1/3] Bluetooth: aosp: surface AOSP quality report through mgmt

From: Joseph Hwang
Date: Sat Mar 05 2022 - 02:42:37 EST

Next message: Joseph Hwang: "Re: [PATCH v4 2/3] Bluetooth: btintel: surface Intel telemetry events through mgmt"
Previous message: David Gow: "[PATCH] list: test: Test the hlist structure"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Marcel, thank you for reviewing the patches! I have some questions
and would like to confirm with you. Please read my replies in the
lines below. Thanks!

On Thu, Feb 17, 2022 at 8:41 PM Marcel Holtmann <marcel@xxxxxxxxxxxx> wrote:
>
> Hi Joseph,
>
> > When receiving a HCI vendor event, the kernel checks if it is an
> > AOSP bluetooth quality report. If yes, the event is sent to bluez
> > user space through the mgmt socket.
> >
> > Signed-off-by: Joseph Hwang <josephsih@xxxxxxxxxxxx>
> > Reviewed-by: Archie Pusaka <apusaka@xxxxxxxxxxxx>
> > ---
> >
> > (no changes since v3)
> >
> > Changes in v3:
> > - Rebase to resolve the code conflict.
> > - Move aosp_quality_report_evt() from hci_event.c to aosp.c.
> > - A new patch (3/3) is added to enable the quality report feature.
> >
> > Changes in v2:
> > - Scrap the two structures defined in aosp.c and use constants for
> > size check.
> > - Do a basic size check about the quality report event. Do not pull
> > data from the event in which the kernel has no interest.
> > - Define vendor event prefixes with which vendor events of distinct
> > vendor specifications can be clearly differentiated.
> > - Use mgmt helpers to add the header and data to a mgmt skb.
>
> this unsolicited vendor event business is giving me a headache. I assume it would be a lot simpler, but it doesn’t look like this. I need to spent some rounds of thinking to give you good advice. Unfortunately since we also want to support Intel vendor specific pieces, this is getting super complicated.

You have mentioned using hdev->hci_recv_quality_report to send the
unsolicited events from the Intel driver in your comments in Patch
2/3. Just would like to confirm with you. Thanks.

>
> > include/net/bluetooth/hci_core.h | 5 ++
> > include/net/bluetooth/mgmt.h | 7 +++
> > net/bluetooth/aosp.c | 27 ++++++++++
> > net/bluetooth/aosp.h | 13 +++++
> > net/bluetooth/hci_event.c | 84 +++++++++++++++++++++++++++++++-
> > net/bluetooth/mgmt.c | 20 ++++++++
> > net/bluetooth/msft.c | 14 ++++++
> > net/bluetooth/msft.h | 12 +++++
> > 8 files changed, 181 insertions(+), 1 deletion(-)
> >
> > diff --git a/include/net/bluetooth/hci_core.h b/include/net/bluetooth/hci_core.h
> > index f5caff1ddb29..ea83619ac4de 100644
> > --- a/include/net/bluetooth/hci_core.h
> > +++ b/include/net/bluetooth/hci_core.h
> > @@ -1864,6 +1864,8 @@ int mgmt_add_adv_patterns_monitor_complete(struct hci_dev *hdev, u8 status);
> > int mgmt_remove_adv_monitor_complete(struct hci_dev *hdev, u8 status);
> > void mgmt_adv_monitor_device_lost(struct hci_dev *hdev, u16 handle,
> > bdaddr_t *bdaddr, u8 addr_type);
> > +int mgmt_quality_report(struct hci_dev *hdev, void *data, u32 data_len,
> > + u8 quality_spec);
> >
> > u8 hci_le_conn_update(struct hci_conn *conn, u16 min, u16 max, u16 latency,
> > u16 to_multiplier);
> > @@ -1882,4 +1884,7 @@ void hci_copy_identity_address(struct hci_dev *hdev, bdaddr_t *bdaddr,
> >
> > #define TRANSPORT_TYPE_MAX 0x04
> >
> > +#define QUALITY_SPEC_AOSP_BQR 0x0
> > +#define QUALITY_SPEC_INTEL_TELEMETRY 0x1
> > +
> > #endif /* __HCI_CORE_H */
> > diff --git a/include/net/bluetooth/mgmt.h b/include/net/bluetooth/mgmt.h
> > index 3d26e6a3478b..83b602636262 100644
> > --- a/include/net/bluetooth/mgmt.h
> > +++ b/include/net/bluetooth/mgmt.h
> > @@ -1120,3 +1120,10 @@ struct mgmt_ev_adv_monitor_device_lost {
> > __le16 monitor_handle;
> > struct mgmt_addr_info addr;
> > } __packed;
> > +
> > +#define MGMT_EV_QUALITY_REPORT 0x0031
> > +struct mgmt_ev_quality_report {
> > + __u8 quality_spec;
> > + __u32 data_len;
> > + __u8 data[];
> > +} __packed;
> > diff --git a/net/bluetooth/aosp.c b/net/bluetooth/aosp.c
> > index 432ae3aac9e3..4a336433180d 100644
> > --- a/net/bluetooth/aosp.c
> > +++ b/net/bluetooth/aosp.c
> > @@ -199,3 +199,30 @@ int aosp_set_quality_report(struct hci_dev *hdev, bool enable)
> > else
> > return disable_quality_report(hdev);
> > }
> > +
> > +#define BLUETOOTH_QUALITY_REPORT_EV 0x58
> > +
> > +/* The following LEN = 1-byte Sub-event code + 48-byte Sub-event Parameters */
> > +#define BLUETOOTH_QUALITY_REPORT_LEN 49
> > +
> > +bool aosp_check_quality_report_len(struct sk_buff *skb)
> > +{
> > + /* skb->len is allowed to be larger than BLUETOOTH_QUALITY_REPORT_LEN
> > + * to accommodate an additional Vendor Specific Parameter (vsp) field.
> > + */
> > + if (skb->len < BLUETOOTH_QUALITY_REPORT_LEN) {
> > + BT_ERR("AOSP evt data len %d too short (%u expected)",
> > + skb->len, BLUETOOTH_QUALITY_REPORT_LEN);
> > + return false;
> > + }
> > +
> > + return true;
> > +}
> > +
> > +void aosp_quality_report_evt(struct hci_dev *hdev, void *data,
> > + struct sk_buff *skb)
> > +{
> > + if (aosp_has_quality_report(hdev) && aosp_check_quality_report_len(skb))
> > + mgmt_quality_report(hdev, skb->data, skb->len,
> > + QUALITY_SPEC_AOSP_BQR);
> > +}
> > diff --git a/net/bluetooth/aosp.h b/net/bluetooth/aosp.h
> > index 2fd8886d51b2..b21751e012de 100644
> > --- a/net/bluetooth/aosp.h
> > +++ b/net/bluetooth/aosp.h
> > @@ -10,6 +10,9 @@ void aosp_do_close(struct hci_dev *hdev);
> >
> > bool aosp_has_quality_report(struct hci_dev *hdev);
> > int aosp_set_quality_report(struct hci_dev *hdev, bool enable);
> > +bool aosp_check_quality_report_len(struct sk_buff *skb);
> > +void aosp_quality_report_evt(struct hci_dev *hdev, void *data,
> > + struct sk_buff *skb);
> >
> > #else
> >
> > @@ -26,4 +29,14 @@ static inline int aosp_set_quality_report(struct hci_dev *hdev, bool enable)
> > return -EOPNOTSUPP;
> > }
> >
> > +static inline bool aosp_check_quality_report_len(struct sk_buff *skb)
> > +{
> > + return false;
> > +}
> > +
> > +static inline void aosp_quality_report_evt(struct hci_dev *hdev, void *data,
> > + struct sk_buff *skb)
> > +{
> > +}
> > +
> > #endif
> > diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
> > index 63b925921c87..6468ea0f71bd 100644
> > --- a/net/bluetooth/hci_event.c
> > +++ b/net/bluetooth/hci_event.c
> > @@ -37,6 +37,7 @@
> > #include "smp.h"
> > #include "msft.h"
> > #include "eir.h"
> > +#include "aosp.h"
> >
> > #define ZERO_KEY "\x00\x00\x00\x00\x00\x00\x00\x00" \
> > "\x00\x00\x00\x00\x00\x00\x00\x00"
> > @@ -4241,6 +4242,87 @@ static void hci_num_comp_blocks_evt(struct hci_dev *hdev, void *data,
> > queue_work(hdev->workqueue, &hdev->tx_work);
> > }
> >
> > +/* Define the fixed vendor event prefixes below.
> > + * Note: AOSP HCI Requirements use 0x54 and up as sub-event codes without
> > + * actually defining a vendor prefix. Refer to
> > + * https://source.android.com/devices/bluetooth/hci_requirements
> > + * Hence, the other vendor event prefixes should not use the same
> > + * space to avoid collision.
> > + */
> > +static unsigned char AOSP_BQR_PREFIX[] = { 0x58 };
> > +
> > +/* Some vendor prefixes are fixed values and lengths. */
> > +#define FIXED_EVT_PREFIX(_prefix, _vendor_func) \
> > +{ \
> > + .prefix = _prefix, \
> > + .prefix_len = sizeof(_prefix), \
> > + .vendor_func = _vendor_func, \
> > + .get_prefix = NULL, \
> > + .get_prefix_len = NULL, \
> > +}
> > +
> > +/* Some vendor prefixes are only available at run time. The
> > + * values and lengths are variable.
> > + */
> > +#define DYNAMIC_EVT_PREFIX(_prefix_func, _prefix_len_func, _vendor_func)\
> > +{ \
> > + .prefix = NULL, \
> > + .prefix_len = 0, \
> > + .vendor_func = _vendor_func, \
> > + .get_prefix = _prefix_func, \
> > + .get_prefix_len = _prefix_len_func, \
> > +}
> > +
> > +/* Every distinct vendor specification must have a well-defined vendor
> > + * event prefix to determine if a vendor event meets the specification.
> > + * If an event prefix is fixed, it should be delcared with FIXED_EVT_PREFIX.
> > + * Otherwise, DYNAMIC_EVT_PREFIX should be used for variable prefixes.
> > + */
> > +struct vendor_event_prefix {
> > + __u8 *prefix;
> > + __u8 prefix_len;
> > + void (*vendor_func)(struct hci_dev *hdev, void *data,
> > + struct sk_buff *skb);
> > + __u8 *(*get_prefix)(struct hci_dev *hdev);
> > + __u8 (*get_prefix_len)(struct hci_dev *hdev);
> > +} evt_prefixes[] = {
> > + FIXED_EVT_PREFIX(AOSP_BQR_PREFIX, aosp_quality_report_evt),
> > + DYNAMIC_EVT_PREFIX(get_msft_evt_prefix, get_msft_evt_prefix_len,
> > + msft_vendor_evt),
> > +
> > + /* end with a null entry */
> > + {},
> > +};
> > +
> > +static void hci_vendor_evt(struct hci_dev *hdev, void *data,
> > + struct sk_buff *skb)
> > +{
> > + int i;
> > + __u8 *prefix;
> > + __u8 prefix_len;
> > +
> > + for (i = 0; evt_prefixes[i].vendor_func; i++) {
> > + if (evt_prefixes[i].get_prefix)
> > + prefix = evt_prefixes[i].get_prefix(hdev);
> > + else
> > + prefix = evt_prefixes[i].prefix;
> > +
> > + if (evt_prefixes[i].get_prefix_len)
> > + prefix_len = evt_prefixes[i].get_prefix_len(hdev);
> > + else
> > + prefix_len = evt_prefixes[i].prefix_len;
> > +
> > + if (!prefix || prefix_len == 0)
> > + continue;
> > +
> > + /* Compare the raw prefix data directly. */
> > + if (!memcmp(prefix, skb->data, prefix_len)) {
> > + evt_prefixes[i].vendor_func(hdev, data, skb);
> > + break;
> > + }
> > + }
> > +}
> > +
> > static void hci_mode_change_evt(struct hci_dev *hdev, void *data,
> > struct sk_buff *skb)
> > {
> > @@ -6844,7 +6926,7 @@ static const struct hci_ev {
> > HCI_EV(HCI_EV_NUM_COMP_BLOCKS, hci_num_comp_blocks_evt,
> > sizeof(struct hci_ev_num_comp_blocks)),
> > /* [0xff = HCI_EV_VENDOR] */
> > - HCI_EV_VL(HCI_EV_VENDOR, msft_vendor_evt, 0, HCI_MAX_EVENT_SIZE),
> > + HCI_EV_VL(HCI_EV_VENDOR, hci_vendor_evt, 0, HCI_MAX_EVENT_SIZE),
> > };
>
> I was thinking along the lines like this:
>
> HCI_EV_VND(evt_prefix, evt_prefix_len, callback),
>
> So that we in the end can do things like this:
>
> HCI_EV_VL({ 0x58 }, 1, aosp_quality_report_evt),
>

I wish I could use a macro as simple as HCI_EV_VND(evt_prefix,
evt_prefix_len, callback). However, I do not have a clean method to
handle the dynamic msft vendor prefix of which the value and length
are not known until runtime. Do you have any suggestions here?

>
>
> >
> > static void hci_event_func(struct hci_dev *hdev, u8 event, struct sk_buff *skb,
> > diff --git a/net/bluetooth/mgmt.c b/net/bluetooth/mgmt.c
> > index 914e2f2d3586..5e48576041fb 100644
> > --- a/net/bluetooth/mgmt.c
> > +++ b/net/bluetooth/mgmt.c
> > @@ -4389,6 +4389,26 @@ static int set_exp_feature(struct sock *sk, struct hci_dev *hdev,
> > MGMT_STATUS_NOT_SUPPORTED);
> > }
> >
> > +int mgmt_quality_report(struct hci_dev *hdev, void *data, u32 data_len,
> > + u8 quality_spec)
> > +{
> > + struct mgmt_ev_quality_report *ev;
> > + struct sk_buff *skb;
> > +
> > + skb = mgmt_alloc_skb(hdev, MGMT_EV_QUALITY_REPORT,
> > + sizeof(*ev) + data_len);
> > + if (!skb)
> > + return -ENOMEM;
> > +
> > + ev = skb_put(skb, sizeof(*ev));
> > + ev->quality_spec = quality_spec;
> > + ev->data_len = data_len;
> > + skb_put_data(skb, data, data_len);
> > +
> > + return mgmt_event_skb(skb, NULL);
> > +}
> > +EXPORT_SYMBOL(mgmt_quality_report);
> > +
>
> I know what you want to do, but I can not let you call mgmt_ function from a driver. We need to make this cleaner and abstract so that the driver has a proper path to report it to hci_core and that decides then to send the report or not.

Will set up hdev->hci_recv_quality_report as you mentioned in comments
in Patch 2/3.

>
>
> > static int get_device_flags(struct sock *sk, struct hci_dev *hdev, void *data,
> > u16 data_len)
> > {
> > diff --git a/net/bluetooth/msft.c b/net/bluetooth/msft.c
> > index 9a3d77d3ca86..3edf64baf479 100644
> > --- a/net/bluetooth/msft.c
> > +++ b/net/bluetooth/msft.c
> > @@ -731,6 +731,20 @@ static void msft_monitor_device_evt(struct hci_dev *hdev, struct sk_buff *skb)
> > handle_data->mgmt_handle);
> > }
> >
> > +__u8 *get_msft_evt_prefix(struct hci_dev *hdev)
> > +{
> > + struct msft_data *msft = hdev->msft_data;
> > +
> > + return msft->evt_prefix;
> > +}
> > +
> > +__u8 get_msft_evt_prefix_len(struct hci_dev *hdev)
> > +{
> > + struct msft_data *msft = hdev->msft_data;
> > +
> > + return msft->evt_prefix_len;
> > +}
> > +
>
> So I wonder if this should be moved directly into hci_dev under CONFIG_BT_MSFTEXT check. Luiz also needs to have a look at this. This is unfortunately getting a bit nasty now. We need to find a clean solution, otherwise the next vendor thing is blowing up in our face.

My understanding is that static inline hci_set_msft_opcode() under
CONFIG_BT_MSFTEXT is placed in hci_core.h because drivers need to call
it. Here, get_msft_evt_prefix() and get_msft_evt_prefix_len() are not
to be called by drivers. Do we still need to move them into
hci_core.h? Please let me know if I have any misunderstanding. Thanks.

>
>
> > void msft_vendor_evt(struct hci_dev *hdev, void *data, struct sk_buff *skb)
> > {
> > struct msft_data *msft = hdev->msft_data;
> > diff --git a/net/bluetooth/msft.h b/net/bluetooth/msft.h
> > index afcaf7d3b1cb..a354ebf61fed 100644
> > --- a/net/bluetooth/msft.h
> > +++ b/net/bluetooth/msft.h
> > @@ -27,6 +27,8 @@ int msft_set_filter_enable(struct hci_dev *hdev, bool enable);
> > int msft_suspend_sync(struct hci_dev *hdev);
> > int msft_resume_sync(struct hci_dev *hdev);
> > bool msft_curve_validity(struct hci_dev *hdev);
> > +__u8 *get_msft_evt_prefix(struct hci_dev *hdev);
> > +__u8 get_msft_evt_prefix_len(struct hci_dev *hdev);
> >
> > #else
> >
> > @@ -77,4 +79,14 @@ static inline bool msft_curve_validity(struct hci_dev *hdev)
> > return false;
> > }
> >
> > +static inline __u8 *get_msft_evt_prefix(struct hci_dev *hdev)
> > +{
> > + return NULL;
> > +}
> > +
> > +static inline __u8 get_msft_evt_prefix_len(struct hci_dev *hdev)
> > +{
> > + return 0;
> > +}
> > +
> > #endif
>
> Regards
>
> Marcel
>

Regards,
--

Joseph Shyh-In Hwang
Email: josephsih@xxxxxxxxxx

Next message: Joseph Hwang: "Re: [PATCH v4 2/3] Bluetooth: btintel: surface Intel telemetry events through mgmt"
Previous message: David Gow: "[PATCH] list: test: Test the hlist structure"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]