Skip to contentSkip to conversation

Bluetooth: btusb: Configure altsetting for USER_CHANNEL

PATCH Superseded Hsin-chen Chuang 5 participants 14 messages Feb 24, 2025 – Mar 3, 2025 Bluetooth drivers

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 Bluetooth: btusb: Configure altsetting for USER_CHANNEL

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +11 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
 	  Say Y here to enable USB poll_sync for Bluetooth USB devices by
 	  default.
 
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+	bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+	depends on BT_HCIBTUSB
+	default n
+	help
+	  Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+	  when SCO connected
+
+	  This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+	  on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
 	bool "Broadcom protocol support"
 	depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +46 -0

70 lines not on this page · raw message

Signature · 1 line
2.48.1.601.g30ceb7b040-goog
#1 lore raw

This is automated email and please do not reply to this email!

Dear submitter,

Thank you for submitting the patches to the linux bluetooth mailing list. This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/project/bluetooth/list/?series=936884

---Test result---

Test Summary:

CheckPatch                    PENDING   0.31 seconds
GitLint                       PENDING   0.19 seconds
SubjectPrefix                 PASS      0.10 seconds
BuildKernel                   PASS      23.90 seconds
CheckAllWarning               PASS      26.56 seconds
CheckSparse                   PASS      31.56 seconds
BuildKernel32                 PASS      23.64 seconds
TestRunnerSetup               PASS      426.38 seconds
TestRunner_l2cap-tester       PASS      20.56 seconds
TestRunner_iso-tester         PASS      37.10 seconds
TestRunner_bnep-tester        PASS      4.74 seconds
TestRunner_mgmt-tester        FAIL      119.03 seconds
TestRunner_rfcomm-tester      PASS      7.90 seconds
TestRunner_sco-tester         PASS      9.50 seconds
TestRunner_ioctl-tester       PASS      8.51 seconds
TestRunner_mesh-tester        PASS      5.80 seconds
TestRunner_smp-tester         PASS      7.12 seconds
TestRunner_userchan-tester    PASS      4.91 seconds
IncrementalBuild              PENDING   0.37 seconds

Details
##############################
Test: CheckPatch - PENDING
Desc: Run checkpatch.pl script Output:

##############################
Test: GitLint - PENDING
Desc: Run gitlint
Output:

##############################
Test: TestRunner_mgmt-tester - FAIL
Desc: Run mgmt-tester with test-runner Output:
Total: 490, Passed: 484 (98.8%), Failed: 2, Not Run: 4

Failed Test Cases

LL Privacy - Add Device 3 (AL is full)               Failed       0.200 seconds
LL Privacy - Set Flags 2 (Enable RL)                 Failed       0.146 seconds

##############################
Test: IncrementalBuild - PENDING
Desc: Incremental build with the patches in the series Output:

---
Regards,
Linux Bluetooth

#2 lore raw
In reply to Hsin-chen Chuang
Quoted 110 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +11 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
 	  Say Y here to enable USB poll_sync for Bluetooth USB devices by
 	  default.
 
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+	bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+	depends on BT_HCIBTUSB
+	default n
+	help
+	  Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+	  when SCO connected
+
+	  This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+	  on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
 	bool "Broadcom protocol support"
 	depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+	IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
 	spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+	struct hci_event_hdr *hdr = (void *) skb->data;
+	struct hci_ev_sync_conn_complete *ev =
+		(void *) skb->data + sizeof(*hdr);
+	struct hci_dev *hdev = data->hdev;
+	unsigned int notify_air_mode;
+
+	if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+		return;
+
+	if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+		return;
+
+	if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+		return;
+
+	switch (ev->air_mode) {
+	case BT_CODEC_CVSD:
+		notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+		break;
+
+	case BT_CODEC_TRANSPARENT:
+		notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+		break;
+
+	default:
+		return;
+	}
+
+	bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+	data->sco_num = 1;
+	data->air_mode = notify_air_mode;
+	schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
 	if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 		schedule_delayed_work(&data->rx_work, 0);
 	}
 
+	/* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+	if (auto_set_isoc_alt &&
+	    hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+		btusb_sco_connected(data, skb);
+
 	return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would not work at all for multiple controllers in teh same system, right? That should cause it to not even be considered at all as a viable solution.

confused,

greg k-h

#3 lore raw

Hi Greg,

Quoted 115 lines · Greg KH
Quoted 113 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +11 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
Quoted 5 lines

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 2 lines
drivers/bluetooth/btusb.c +2 -0
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Sorry but could you please provide an example to guard a feature like this.

Quoted 1 line

not work at all for multiple controllers in teh same system, right?

Do you mean we can't have separate parameters for different controllers? Yes that's true, but why would a user want the different behavior on the same machine?

Quoted 6 lines

That should cause it to not even be considered at all as a viable solution.

confused,

greg k-h

Signature · 2 lines
Best Regards,
Hsin-chen
#4 lore raw
Quoted 118 lines · Hsin-chen Chuang

Hi Greg,

Quoted 115 lines · Greg KH
Quoted 113 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +11 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
Quoted 7 lines
Quoted 5 lines

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 2 lines
drivers/bluetooth/btusb.c +2 -0
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain to configure, we have loads of other ways to do this now (configfs, debugfs for debugging stuff, sysfs for device-specific things, etc.)

Quoted 5 lines
Quoted 1 line

not work at all for multiple controllers in teh same system, right?

Do you mean we can't have separate parameters for different controllers? Yes that's true, but why would a user want the different behavior on the same machine?

Why would you prevent them from allowing this to happen for a device-specific option?

thanks,

greg k-h

#5 lore raw
Quoted 126 lines · Greg KH
Quoted 124 lines · Hsin-chen Chuang

Hi Greg,

Quoted 120 lines · Greg KH
Quoted 110 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +11 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Quoted 15 lines

to configure, we have loads of other ways to do this now (configfs, debugfs for debugging stuff, sysfs for device-specific things, etc.)

Quoted 5 lines
Quoted 1 line

not work at all for multiple controllers in teh same system, right?

Do you mean we can't have separate parameters for different controllers? Yes that's true, but why would a user want the different behavior on the same machine?

Why would you prevent them from allowing this to happen for a device-specific option?

thanks,

greg k-h

Signature · 2 lines
Best Regards,
Hsin-chen
#6 lore raw

Hi Hsin-chen,

Quoted 36 lines · Hsin-chen Chuang
Quoted 34 lines · Greg KH
Quoted 32 lines · Hsin-chen Chuang

Hi Greg,

Quoted 29 lines · Greg KH
Quoted 27 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +0 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/bluetooth/Kconfig config BT_HCIBTUSB_POLL_SYNC · line 62

Quoted 4 lines
drivers/bluetooth/Kconfig +4 -0
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n

Maybe we can do just:

default y if CHROME_PLATFORMS

Quoted 93 lines
Quoted 88 lines
Quoted 88 lines
Quoted 87 lines
Quoted 79 lines
drivers/bluetooth/Kconfig +7 -0
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

Quoted 8 lines

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Perhaps you can just use CHROME_PLATFORMS (suggested above) in Kconfig to enable intercepting of the events, etc, so we don't need any runtime parameters.

Quoted 19 lines
Quoted 9 lines

to configure, we have loads of other ways to do this now (configfs, debugfs for debugging stuff, sysfs for device-specific things, etc.)

Quoted 5 lines
Quoted 1 line

not work at all for multiple controllers in teh same system, right?

Do you mean we can't have separate parameters for different controllers? Yes that's true, but why would a user want the different behavior on the same machine?

Quoted 6 lines

Why would you prevent them from allowing this to happen for a device-specific option?

thanks,

greg k-h

--
Best Regards,
Hsin-chen

Signature · 1 line
Luiz Augusto von Dentz
#7 lore raw

Hi Luiz,

Quoted 151 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 36 lines · Hsin-chen Chuang
Quoted 34 lines · Greg KH
Quoted 32 lines · Hsin-chen Chuang

Hi Greg,

Quoted 29 lines · Greg KH
Quoted 27 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +0 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/bluetooth/Kconfig config BT_HCIBTUSB_POLL_SYNC · line 62

Quoted 4 lines
drivers/bluetooth/Kconfig +4 -0
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n

Maybe we can do just:

default y if CHROME_PLATFORMS

Quoted 93 lines
Quoted 88 lines
Quoted 88 lines
Quoted 87 lines
Quoted 79 lines
drivers/bluetooth/Kconfig +7 -0
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

Quoted 8 lines

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Quoted 3 lines

Perhaps you can just use CHROME_PLATFORMS (suggested above) in Kconfig to enable intercepting of the events, etc, so we don't need any runtime parameters.

I'm afraid that this doesn't resolve Greg's comment below because the multiple controllers are still bonded to the same config. Also I would hesitate to put this Chrome devices specific because project Floss shall be able to run on a general Linux environment. If you have a strong opinion to guard this behind a flag, perhaps we could try other options suggested by Greg (configfs or maybe back to sysfs).

Quoted 17 lines
Quoted 17 lines
Quoted 16 lines

to configure, we have loads of other ways to do this now (configfs, debugfs for debugging stuff, sysfs for device-specific things, etc.)

Quoted 5 lines
Quoted 1 line

not work at all for multiple controllers in teh same system, right?

Do you mean we can't have separate parameters for different controllers? Yes that's true, but why would a user want the different behavior on the same machine?

Why would you prevent them from allowing this to happen for a device-specific option?

thanks,

greg k-h

Quoted 8 lines
Quoted 3 lines

--
Best Regards,
Hsin-chen

--
Luiz Augusto von Dentz

Best Regards,

Hsin-chen

#8 lore raw

Hi Hsin-chen,

Quoted 156 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 151 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 36 lines · Hsin-chen Chuang
Quoted 34 lines · Greg KH
Quoted 32 lines · Hsin-chen Chuang

Hi Greg,

Quoted 29 lines · Greg KH
Quoted 27 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +0 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/bluetooth/Kconfig config BT_HCIBTUSB_POLL_SYNC · line 62

Quoted 4 lines
drivers/bluetooth/Kconfig +4 -0
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n

Maybe we can do just:

default y if CHROME_PLATFORMS

Quoted 101 lines
Quoted 88 lines
Quoted 88 lines
Quoted 87 lines
Quoted 79 lines
drivers/bluetooth/Kconfig +7 -0
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Quoted 6 lines
Quoted 3 lines

Perhaps you can just use CHROME_PLATFORMS (suggested above) in Kconfig to enable intercepting of the events, etc, so we don't need any runtime parameters.

I'm afraid that this doesn't resolve Greg's comment below because the multiple controllers are still bonded to the same config.

Well that would be enabled for every controller plugged into the system.

Quoted 2 lines

Also I would hesitate to put this Chrome devices specific because project Floss shall be able to run on a general Linux environment.

It can be enabled for Linux in general, it just not enabled by default which can be changed later if someone decides to use Floss outside of Chrome OS, that said I very much doubt Google would be supporting anything other then Chrome based platforms or Floss don't depend on Android vendor HCI commands? At least the Android documentation seems to suggest a bunch of vendor commands are required:

https://source.android.com/docs/core/connect/bluetooth/hci_requirements

Quoted 3 lines

If you have a strong opinion to guard this behind a flag, perhaps we could try other options suggested by Greg (configfs or maybe back to sysfs).

I think those would be more complicated then handling this via Kconfig, besides I think this works best since there is no new API introduced, just a new Kconfig option which can be enabled in any platform/distro.

Quoted 30 lines
Quoted 17 lines
Quoted 17 lines
Quoted 16 lines

to configure, we have loads of other ways to do this now (configfs, debugfs for debugging stuff, sysfs for device-specific things, etc.)

Quoted 5 lines
Quoted 1 line

not work at all for multiple controllers in teh same system, right?

Do you mean we can't have separate parameters for different controllers? Yes that's true, but why would a user want the different behavior on the same machine?

Why would you prevent them from allowing this to happen for a device-specific option?

thanks,

greg k-h

Quoted 8 lines
Quoted 3 lines

--
Best Regards,
Hsin-chen

--
Luiz Augusto von Dentz

Best Regards,

Hsin-chen

Signature · 1 line
Luiz Augusto von Dentz
#9 lore raw

Hi Luiz,

Quoted 160 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 156 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 151 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 36 lines · Hsin-chen Chuang
Quoted 34 lines · Greg KH
Quoted 32 lines · Hsin-chen Chuang

Hi Greg,

Quoted 29 lines · Greg KH
Quoted 27 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +0 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/bluetooth/Kconfig config BT_HCIBTUSB_POLL_SYNC · line 62

Quoted 4 lines
drivers/bluetooth/Kconfig +4 -0
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n

Maybe we can do just:

default y if CHROME_PLATFORMS

Quoted 101 lines
Quoted 88 lines
Quoted 88 lines
Quoted 87 lines
Quoted 79 lines
drivers/bluetooth/Kconfig +7 -0
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

Perhaps you can just use CHROME_PLATFORMS (suggested above) in Kconfig to enable intercepting of the events, etc, so we don't need any runtime parameters.

I'm afraid that this doesn't resolve Greg's comment below because the multiple controllers are still bonded to the same config.

Well that would be enabled for every controller plugged into the system.

To clarify, I'm totally fine with runtime parameters and/or any Kconfig like CHROME_PLATFORMS, although I'd prefer the current patch more.

But I guess Greg's point is that the multiple controllers should be able to have different configurations at the same time. I'd respect your decision to accept this patch or another patch with CHROME_PLATFORMS, but I guess it's better to convince Greg first.

Quoted 9 lines
Quoted 3 lines

Also I would hesitate to put this Chrome devices specific because project Floss shall be able to run on a general Linux environment.

It can be enabled for Linux in general, it just not enabled by default which can be changed later if someone decides to use Floss outside of Chrome OS, that said I very much doubt Google would be supporting anything other then Chrome based platforms or Floss don't depend on Android vendor HCI commands? At least the Android documentation seems

Floss doesn't depend on Android extensions - Our hardwares doesn't support that either :p

Quoted 40 lines

to suggest a bunch of vendor commands are required:

https://source.android.com/docs/core/connect/bluetooth/hci_requirements

Quoted 3 lines

If you have a strong opinion to guard this behind a flag, perhaps we could try other options suggested by Greg (configfs or maybe back to sysfs).

I think those would be more complicated then handling this via Kconfig, besides I think this works best since there is no new API introduced, just a new Kconfig option which can be enabled in any platform/distro.

Quoted 27 lines
Quoted 25 lines
Quoted 13 lines
Quoted 13 lines

to configure, we have loads of other ways to do this now (configfs, debugfs for debugging stuff, sysfs for device-specific things, etc.)

Quoted 5 lines
Quoted 1 line

not work at all for multiple controllers in teh same system, right?

Do you mean we can't have separate parameters for different controllers? Yes that's true, but why would a user want the different behavior on the same machine?

Why would you prevent them from allowing this to happen for a device-specific option?

Quoted 7 lines
Quoted 3 lines

thanks,

greg k-h

--
Best Regards,
Hsin-chen

--
Luiz Augusto von Dentz

Quoted 8 lines
Quoted 3 lines

Best Regards,

Hsin-chen

--
Luiz Augusto von Dentz

Best Regards,

Hsin-chen

#10 lore raw

Hi Hsin-chen,

Quoted 174 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 160 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 156 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 151 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 36 lines · Hsin-chen Chuang
Quoted 34 lines · Greg KH
Quoted 32 lines · Hsin-chen Chuang

Hi Greg,

Quoted 29 lines · Greg KH
Quoted 27 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +0 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/bluetooth/Kconfig config BT_HCIBTUSB_POLL_SYNC · line 62

Quoted 4 lines
drivers/bluetooth/Kconfig +4 -0
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n

Maybe we can do just:

default y if CHROME_PLATFORMS

Quoted 101 lines
Quoted 88 lines
Quoted 88 lines
Quoted 87 lines
Quoted 79 lines
drivers/bluetooth/Kconfig +7 -0
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

Perhaps you can just use CHROME_PLATFORMS (suggested above) in Kconfig to enable intercepting of the events, etc, so we don't need any runtime parameters.

I'm afraid that this doesn't resolve Greg's comment below because the multiple controllers are still bonded to the same config.

Well that would be enabled for every controller plugged into the system.

Quoted 8 lines

To clarify, I'm totally fine with runtime parameters and/or any Kconfig like CHROME_PLATFORMS, although I'd prefer the current patch more.

But I guess Greg's point is that the multiple controllers should be able to have different configurations at the same time. I'd respect your decision to accept this patch or another patch with CHROME_PLATFORMS, but I guess it's better to convince Greg first.

Perhaps I got Greg's response wrong, but I think he was suggesting that it should just work without any special configuration, which is what I'm suggesting here, the use of CHROME_PLATFORMS is more of a stopgap until proper test automation is introduced.

You can already have different configurations with respect to HCI_CHANNEL_USER since that is done per controller, btw that is used by the likes of btproxy tool for example which is normally used to hook the controller into a VM.

#11 lore raw

Hi Luiz,

Quoted 192 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 174 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 160 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 156 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 151 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 36 lines · Hsin-chen Chuang
Quoted 34 lines · Greg KH
Quoted 32 lines · Hsin-chen Chuang

Hi Greg,

Quoted 29 lines · Greg KH
Quoted 27 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +0 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/bluetooth/Kconfig config BT_HCIBTUSB_POLL_SYNC · line 62

Quoted 4 lines
drivers/bluetooth/Kconfig +4 -0
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n

Maybe we can do just:

default y if CHROME_PLATFORMS

Quoted 101 lines
Quoted 88 lines
Quoted 88 lines
Quoted 87 lines
Quoted 79 lines
drivers/bluetooth/Kconfig +7 -0
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

Perhaps you can just use CHROME_PLATFORMS (suggested above) in Kconfig to enable intercepting of the events, etc, so we don't need any runtime parameters.

I'm afraid that this doesn't resolve Greg's comment below because the multiple controllers are still bonded to the same config.

Well that would be enabled for every controller plugged into the system.

Quoted 8 lines

To clarify, I'm totally fine with runtime parameters and/or any Kconfig like CHROME_PLATFORMS, although I'd prefer the current patch more.

But I guess Greg's point is that the multiple controllers should be able to have different configurations at the same time. I'd respect your decision to accept this patch or another patch with CHROME_PLATFORMS, but I guess it's better to convince Greg first.

Perhaps I got Greg's response wrong, but I think he was suggesting that it should just work without any special configuration, which is what I'm suggesting here, the use of CHROME_PLATFORMS is more of a stopgap until proper test automation is introduced.

Quoted 4 lines

You can already have different configurations with respect to HCI_CHANNEL_USER since that is done per controller, btw that is used by the likes of btproxy tool for example which is normally used to hook the controller into a VM.

Thanks for the explanation. I'll follow the suggestion in v3.

Best Regards,

Hsin-chen

#12 lore raw

Beyond power concerns, another potential issue is USB host controller scheduling fairness. SCO data is transmitted over isochronous pipes, and if the alternate setting remains non-zero, the USB host controller reserves bandwidth even when no audio data is being transmitted. This could reduce throughput for some USB 1.x/2.0 applications, such as file transfers from USB mass storage devices, when sharing the bus with a Bluetooth controller in SCO mode. I don’t have a Chromebook to verify this theory, but I suggest the ChromeOS team measure the potential impact and follow up if necessary.

Quoted 208 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 192 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 174 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 160 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 156 lines · Hsin-chen Chuang

Hi Luiz,

Quoted 151 lines · Luiz Augusto von Dentz

Hi Hsin-chen,

Quoted 36 lines · Hsin-chen Chuang
Quoted 34 lines · Greg KH
Quoted 32 lines · Hsin-chen Chuang

Hi Greg,

Quoted 29 lines · Greg KH
Quoted 27 lines · Hsin-chen Chuang

From: Hsin-chen Chuang <[email protected]>

Automatically configure the altsetting for USER_CHANNEL when a SCO is connected. This adds support for the USER_CHANNEL to transfer SCO data over USB transport.

Fixes b16b327edb4d ("Bluetooth: btusb: add sysfs attribute to control USB alt setting")

---

Changes in v2:
- Give up tracking the SCO handles. Only configure the altsetting when SCO connected.
- Put the change behind Kconfig/module parameter

 drivers/bluetooth/Kconfig | 11 ++++++++++
 drivers/bluetooth/btusb.c | 46 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+)
drivers/bluetooth/Kconfig +0 -0
index 4ab32abf0f48..7c497f878732 100644
@@ -56,6 +56,17 @@ config BT_HCIBTUSB_POLL_SYNC
        Say Y here to enable USB poll_sync for Bluetooth USB devices by
        default.
 
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

drivers/bluetooth/Kconfig config BT_HCIBTUSB_POLL_SYNC · line 62

Quoted 4 lines
drivers/bluetooth/Kconfig +4 -0
+config BT_HCIBTUSB_AUTO_SET_ISOC_ALT
+     bool "Auto set isoc_altsetting for USER_CHANNEL when SCO connected"
+     depends on BT_HCIBTUSB
+     default n

Maybe we can do just:

default y if CHROME_PLATFORMS

Quoted 101 lines
Quoted 88 lines
Quoted 88 lines
Quoted 87 lines
Quoted 79 lines
drivers/bluetooth/Kconfig +7 -0
+     help
+       Say Y here to enable auto set isoc_altsetting for USER_CHANNEL
+       when SCO connected
+
+       This can be overridden by passing btusb.auto_set_isoc_alt=[y|n]
+       on the kernel commandline.
+
 config BT_HCIBTUSB_BCM
      bool "Broadcom protocol support"
      depends on BT_HCIBTUSB
drivers/bluetooth/btusb.c +43 -0
index de3fa725d210..af93d757911b 100644
@@ -34,6 +34,8 @@ static bool force_scofix;
 static bool enable_autosuspend = IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTOSUSPEND);
 static bool enable_poll_sync = IS_ENABLED(CONFIG_BT_HCIBTUSB_POLL_SYNC);
 static bool reset = true;
+static bool auto_set_isoc_alt =
+     IS_ENABLED(CONFIG_BT_HCIBTUSB_AUTO_SET_ISOC_ALT);
 
 static struct usb_driver btusb_driver;
 
@@ -1113,6 +1115,42 @@ static inline void btusb_free_frags(struct btusb_data *data)
      spin_unlock_irqrestore(&data->rxlock, flags);
 }
 
+static void btusb_sco_connected(struct btusb_data *data, struct sk_buff *skb)
+{
+     struct hci_event_hdr *hdr = (void *) skb->data;
+     struct hci_ev_sync_conn_complete *ev =
+             (void *) skb->data + sizeof(*hdr);
+     struct hci_dev *hdev = data->hdev;
+     unsigned int notify_air_mode;
+
+     if (hci_skb_pkt_type(skb) != HCI_EVENT_PKT)
+             return;
+
+     if (skb->len < sizeof(*hdr) || hdr->evt != HCI_EV_SYNC_CONN_COMPLETE)
+             return;
+
+     if (skb->len != sizeof(*hdr) + sizeof(*ev) || ev->status)
+             return;
+
+     switch (ev->air_mode) {
+     case BT_CODEC_CVSD:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_CVSD;
+             break;
+
+     case BT_CODEC_TRANSPARENT:
+             notify_air_mode = HCI_NOTIFY_ENABLE_SCO_TRANSP;
+             break;
+
+     default:
+             return;
+     }
+
+     bt_dev_info(hdev, "enabling SCO with air mode %u", ev->air_mode);
+     data->sco_num = 1;
+     data->air_mode = notify_air_mode;
+     schedule_work(&data->work);
+}
+
 static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
 {
      if (data->intr_interval) {
@@ -1120,6 +1158,11 @@ static int btusb_recv_event(struct btusb_data *data, struct sk_buff *skb)
              schedule_delayed_work(&data->rx_work, 0);
      }
 
+     /* Configure altsetting for HCI_USER_CHANNEL on SCO connected */
+     if (auto_set_isoc_alt &&
+         hci_dev_test_flag(data->hdev, HCI_USER_CHANNEL))
+             btusb_sco_connected(data, skb);
+
      return data->recv_event(data->hdev, skb);
 }
 

drivers/bluetooth/btusb.c MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default"); · line 4401

Quoted 5 lines
drivers/bluetooth/btusb.c +2 -0
@@ -4354,6 +4397,9 @@ MODULE_PARM_DESC(enable_autosuspend, "Enable USB autosuspend by default");
 module_param(reset, bool, 0644);
 MODULE_PARM_DESC(reset, "Send HCI reset command on initialization");
 
+module_param(auto_set_isoc_alt, bool, 0644);
+MODULE_PARM_DESC(auto_set_isoc_alt, "Auto set isoc_altsetting for USER_CHANNEL when SCO connected");

This is not the 1990's, why are you adding new module parameters when we have so many other more proper ways to do this? And really, this would

Quoted 4 lines
Quoted 1 line

Sorry but could you please provide an example to guard a feature like this.

Depends on what you want to do with this configuration. Why is it an option at all? Why can't it "just work"? Module parameters are a pain

I would like to hand this question to Luiz. I believe this patch just works because this configuration is defined in the spec.
I think Luiz's point is to project the potential existing user, but there's probably no User channel user sending SCO data with the latest btusb driver because:
a) There's no way to configure alt setting from userspace
b) Before eafcfcfca97d, SCO data would be rejected since User channel shouldn't be able to modify hci_conn_num

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

Perhaps you can just use CHROME_PLATFORMS (suggested above) in Kconfig to enable intercepting of the events, etc, so we don't need any runtime parameters.

I'm afraid that this doesn't resolve Greg's comment below because the multiple controllers are still bonded to the same config.

Well that would be enabled for every controller plugged into the system.

Quoted 8 lines

To clarify, I'm totally fine with runtime parameters and/or any Kconfig like CHROME_PLATFORMS, although I'd prefer the current patch more.

But I guess Greg's point is that the multiple controllers should be able to have different configurations at the same time. I'd respect your decision to accept this patch or another patch with CHROME_PLATFORMS, but I guess it's better to convince Greg first.

Perhaps I got Greg's response wrong, but I think he was suggesting that it should just work without any special configuration, which is what I'm suggesting here, the use of CHROME_PLATFORMS is more of a stopgap until proper test automation is introduced.

Quoted 4 lines

You can already have different configurations with respect to HCI_CHANNEL_USER since that is done per controller, btw that is used by the likes of btproxy tool for example which is normally used to hook the controller into a VM.

Thanks for the explanation. I'll follow the suggestion in v3.

Best Regards,

Hsin-chen

#13 lore raw
Quoted 10 lines · Ying Hsu

Beyond power concerns, another potential issue is USB host controller scheduling fairness. SCO data is transmitted over isochronous pipes, and if the alternate setting remains non-zero, the USB host controller reserves bandwidth even when no audio data is being transmitted. This could reduce throughput for some USB 1.x/2.0 applications, such as file transfers from USB mass storage devices, when sharing the bus with a Bluetooth controller in SCO mode. I don’t have a Chromebook to verify this theory, but I suggest the ChromeOS team measure the potential impact and follow up if necessary.

Thanks for the input. We will measure the potential impact.