Bluetooth: hci_bcm: Add wake-up and PM runtime support
Subsystems (2)
v5 cover Bluetooth: hci_bcm: Add wake-up and PM runtime support
Add wake-up capabilities by retrieveing interruption used by BCM device in ACPI table.
Add PM runtime support.
v4->v5:
- Add bt_warn to to unify Bluetooth traces in BT_WARN
- Change bcm_suspend_device/bcm_resume_device prototype to be usable directly in SET_RUNTIME_PM_OPS. This allows to remove bcm_runtime_suspend/bcm_runtime_resume which just call bcm_suspend_device/bcm_resume_device.
- Update commit message and add comment in hci_uart_tty_receive to explain why spin_lock is not requested
v3->v4:
- Add new BT_WARN and bt_dev_warn logging macros
- Use bt_dev_warn for T100 irq polarity fix
- Rename __bcm_suspend to bcm_suspend_device,
and __bcm_resume to bcm_resume_device
- Remove useless rx_lock spinlock. It is only used in hci_uart_tty_receive() which is the receive_buf ldisc callback. This is already protected by flush_to_ldisc() in drivers/tty/tty_buffer.c.
This allows to use mutex in bcm_recv() and call pm_runtime_* directly
- Remove pm_work Work queue which is now useless
- Delay auto-suspend only when reception of a packet is completed
v2->v3:
- Use DMI_EXACT_MATCH instead of DMI_MATCH
- Fix IRQ polarity for T100TA in driver_data of dmi_system_id struct
- Use dmi_first_match() instead of dmi_check_system()
v1->v2:
- Split 1st patch between general wake-up capability and T100TA IRQ fix
- Replace multiple "if ... else if" by switch in bcm_resource()
- Move code to limit number of #ifdef
- Use DMI info to restrict IRQ to T100TA
- Split 2nd patch to prepare PM runtime support in separated patch
- Tested with and without CONFIG_PM_SLEEP and CONFIG_PM.
Frederic Danis (5):
Bluetooth: Add BT_WARN and bt_dev_warn logging macros
Bluetooth: hci_bcm: Fix IRQ polarity for T100
Bluetooth: hci_bcm: Prepare PM runtime support
Bluetooth: Remove useless rx_lock spinlock
Bluetooth: hci_bcm: Add suspend/resume runtime PM functions
drivers/bluetooth/hci_bcm.c | 171 +++++++++++++++++++++++++++++++------- drivers/bluetooth/hci_ldisc.c | 8 +- drivers/bluetooth/hci_uart.h | 1 - include/net/bluetooth/bluetooth.h | 5 ++ net/bluetooth/lib.c | 16 ++++ 5 files changed, 164 insertions(+), 37 deletions(-)
Signature · 1 line
1.9.1
PATCH v5 1/5 Bluetooth: Add BT_WARN and bt_dev_warn logging macros
Add warning logging macros to bluetooth subsystem logs.
--- include/net/bluetooth/bluetooth.h | 5 +++++ net/bluetooth/lib.c | 16 ++++++++++++++++ 2 files changed, 21 insertions(+)
include/net/bluetooth/bluetooth.h +5 -0
@@ -122,11 +122,14 @@ struct bt_voice { __printf(1, 2)void bt_info (const char *fmt, ...); __printf(1, 2) +void bt_warn (const char *fmt, ...); +__printf(1, 2)void bt_err (const char *fmt, ...); __printf(1, 2)void bt_err_ratelimited (const char *fmt, ...);#define BT_INFO (fmt, ...) bt_info(fmt "\n", ##__VA_ARGS__) +#define BT_WARN (fmt, ...) bt_warn(fmt "\n", ##__VA_ARGS__)#define BT_ERR (fmt, ...) bt_err(fmt "\n", ##__VA_ARGS__)#define BT_DBG (fmt, ...) pr_debug(fmt "\n", ##__VA_ARGS__)
@@ -134,6 +137,8 @@ void bt_err_ratelimited(const char *fmt, ...);#define bt_dev_info (hdev, fmt, ...) \ BT_INFO("%s: " fmt, (hdev)->name, ##__VA_ARGS__) +#define bt_dev_warn (hdev, fmt, ...) \ + BT_WARN("%s: " fmt, (hdev)->name, ##__VA_ARGS__)#define bt_dev_err (hdev, fmt, ...) \ BT_ERR("%s: " fmt, (hdev)->name, ##__VA_ARGS__)#define bt_dev_dbg (hdev, fmt, ...) \
net/bluetooth/lib.c +16 -0
@@ -151,6 +151,22 @@ void bt_info(const char *format, ...) }EXPORT_SYMBOL (bt_info); +void bt_warn (const char *format, ...) +{ +struct va_format vaf; +va_list args; + +va_start (args, format); + + vaf.fmt = format; + vaf.va = &args; + +pr_warn ("%pV" , &vaf); + +va_end (args); +} +EXPORT_SYMBOL (bt_warn); +void bt_err (const char *format, ...) {struct va_format vaf;
Signature · 1 line
1.9.1
PATCH v5 2/5 Bluetooth: hci_bcm: Fix IRQ polarity for T100
In reply to Frederic DanisACPI table for BCM2E39 of T100TA is not correct.
Set correct irq_polarity for this device.
--- drivers/bluetooth/hci_bcm.c | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+)
drivers/bluetooth/hci_bcm.c +25 -0
@@ -32,6 +32,7 @@#include <linux/gpio/consumer.h> #include <linux/tty.h> #include <linux/interrupt.h> +#include <linux/dmi.h> #include <net/bluetooth/bluetooth.h> #include <net/bluetooth/hci_core.h>
@@ -513,6 +514,22 @@ static const struct acpi_gpio_mapping acpi_bcm_default_gpios[] = { };#ifdef CONFIG_ACPI +static u8 acpi_active_low =ACPI_ACTIVE_LOW ; + +/* IRQ polarity of some chipsets are not defined correctly in ACPI table. */ +static const struct dmi_system_id bcm_wrong_irq_dmi_table[] = { + { + .ident ="Asus T100TA" , + .matches = { +DMI_EXACT_MATCH (DMI_SYS_VENDOR , +"ASUSTeK COMPUTER INC." ), +DMI_EXACT_MATCH (DMI_PRODUCT_NAME ,"T100TA" ), + }, + .driver_data = &acpi_active_low, + }, + { } +}; +static int bcm_resource (struct acpi_resource *ares,void *data) {struct bcm_device *dev = data;
@@ -552,6 +569,7 @@ static int bcm_acpi_probe(struct bcm_device *dev)const struct acpi_device_id *id;struct acpi_device *adev;LIST_HEAD (resources); +const struct dmi_system_id *dmi_id;int ret; id =acpi_match_device (pdev->dev.driver->acpi_match_table, &pdev->dev);
@@ -608,6 +626,13 @@ static int bcm_acpi_probe(struct bcm_device *dev)acpi_dev_get_resources (adev, &resources, bcm_resource, dev); + dmi_id =dmi_first_match (bcm_wrong_irq_dmi_table); +if (dmi_id) { +bt_dev_warn (dev,"%s: Overwriting IRQ polarity to active low" , + dmi_id->ident); + dev->irq_polarity = *(u8 *)dmi_id->driver_data; + } + return 0; }#else
Signature · 1 line
1.9.1
PATCH v5 3/5 Bluetooth: hci_bcm: Prepare PM runtime support
In reply to Frederic DanisChange some CONFIG_PM_SLEEP to CONFIG_PM as hu and is_suspended parameters will be used during PM runtime callbacks.
Add bcm_suspend_device() and bcm_resume_device() which performs link management for PM callbacks.
These functions will be used for runtime management.
--- drivers/bluetooth/hci_bcm.c | 88 ++++++++++++++++++++++++++++++--------------- 1 file changed, 60 insertions(+), 28 deletions(-)
drivers/bluetooth/hci_bcm.c +60 -28
@@ -56,7 +56,7 @@ struct bcm_device {int irq;u8 irq_polarity; -#ifdef CONFIG_PM_SLEEP +#ifdef CONFIG_PM struct hci_uart *hu;bool is_suspended;/* suspend/resume flag */ #endif
@@ -153,7 +153,7 @@ static int bcm_gpio_set_power(struct bcm_device *dev, bool powered)return 0 ; } -#ifdef CONFIG_PM_SLEEP +#ifdef CONFIG_PM static irqreturn_t bcm_host_wake (int irq,void *data) {struct bcm_device *bdev = data;
@@ -259,7 +259,7 @@ static int bcm_open(struct hci_uart *hu)if (hu->tty->dev->parent == dev->pdev->dev.parent) { bcm->dev = dev; hu->init_speed = dev->init_speed; -#ifdef CONFIG_PM_SLEEP +#ifdef CONFIG_PM dev->hu = hu;#endif bcm_gpio_set_power (bcm->dev, true);
@@ -283,7 +283,7 @@ static int bcm_close(struct hci_uart *hu)mutex_lock (&bcm_device_lock);if (bcm_device_exists (bdev)) {bcm_gpio_set_power (bdev, false); -#ifdef CONFIG_PM_SLEEP +#ifdef CONFIG_PM if (device_can_wakeup (&bdev->pdev->dev)) {devm_free_irq (&bdev->pdev->dev, bdev->irq, bdev);device_init_wakeup (&bdev->pdev->dev, false);
@@ -425,24 +425,17 @@ static struct sk_buff *bcm_dequeue(struct hci_uart *hu)return skb_dequeue (&bcm->txq); } -#ifdef CONFIG_PM_SLEEP -/* Platform suspend callback */ -static int bcm_suspend (struct device *dev) +#ifdef CONFIG_PM +static int bcm_suspend_device (struct device *dev) {struct bcm_device *bdev =platform_get_drvdata (to_platform_device (dev)); -int error; -bt_dev_dbg (bdev,"suspend: is_suspended %d" , bdev->is_suspended); - -mutex_lock (&bcm_device_lock); - -if (!bdev->hu) - goto unlock; +bt_dev_dbg (bdev,"" ); -if (!bdev->is_suspended) { +if (!bdev->is_suspended && bdev->hu) {hci_uart_set_flow_control (bdev->hu, true); -/* Once this callback returns, driver suspends BT via GPIO */ +/* Once this returns, driver suspends BT via GPIO */ bdev->is_suspended = true; }
@@ -453,6 +446,52 @@ static int bcm_suspend(struct device *dev) mdelay(15); } +return 0 ; +} + +static int bcm_resume_device (struct device *dev) +{ +struct bcm_device *bdev =platform_get_drvdata (to_platform_device (dev)); + +bt_dev_dbg (bdev,"" ); + +if (bdev->device_wakeup) { +gpiod_set_value (bdev->device_wakeup, true); +bt_dev_dbg (bdev,"resume, delaying 15 ms" ); +mdelay (15 ); + } + +/* When this executes, the device has woken up already */ +if (bdev->is_suspended && bdev->hu) { + bdev->is_suspended = false; + +hci_uart_set_flow_control (bdev->hu, false); + } + +return 0 ; +} +#endif + +#ifdef CONFIG_PM_SLEEP +/* Platform suspend callback */ +static int bcm_suspend (struct device *dev) +{ +struct bcm_device *bdev =platform_get_drvdata (to_platform_device (dev)); +int error; + +bt_dev_dbg (bdev,"suspend: is_suspended %d" , bdev->is_suspended); + +/* bcm_suspend can be called at any time as long as platform device is +* bound, so it should use bcm_device_lock to protect access to hci_uart +* and device_wake-up GPIO. +*/ +mutex_lock (&bcm_device_lock); + +if (!bdev->hu) + goto unlock; + +bcm_suspend_device (dev); +if (device_may_wakeup (&bdev->pdev->dev)) { error =enable_irq_wake (bdev->irq);if (!error)
@@ -472,6 +511,10 @@ static int bcm_resume(struct device *dev)bt_dev_dbg (bdev,"resume: is_suspended %d" , bdev->is_suspended); +/* bcm_resume can be called at any time as long as platform device is +* bound, so it should use bcm_device_lock to protect access to hci_uart +* and device_wake-up GPIO. +*/ mutex_lock (&bcm_device_lock);if (!bdev->hu)
@@ -482,18 +525,7 @@ static int bcm_resume(struct device *dev)bt_dev_dbg (bdev,"BCM irq: disabled" ); } -if (bdev->device_wakeup) { -gpiod_set_value (bdev->device_wakeup, true); -bt_dev_dbg (bdev,"resume, delaying 15 ms" ); -mdelay (15 ); - } - -/* When this callback executes, the device has woken up already */ -if (bdev->is_suspended) { - bdev->is_suspended = false; - -hci_uart_set_flow_control (bdev->hu, false); - } +bcm_resume_device (dev); unlock:mutex_unlock (&bcm_device_lock);
Signature · 1 line
1.9.1
PATCH v5 4/5 Bluetooth: Remove useless rx_lock spinlock
In reply to Frederic Danisrx_lock spinlock is only used in hci_uart_tty_receive() which is the receive_buf ldisc callback.
hci_uart_tty_receive() is protected from re-entrance by its only caller (flush_to_ldisc() in drivers/tty/tty_buffer.c) which held a mutex (buf->lock) for this section.
This lock allows "safe use of the line discipline's receive_buf() method by excluding the buffer work and any pending flush from using the flip buffer." (comments from tty_buffer_lock_exclusive() in drivers/tty/tty_buffer.c)
So, no need to double protect this resource with rx_lock.
--- drivers/bluetooth/hci_ldisc.c | 8 +++----- drivers/bluetooth/hci_uart.h | 1 - 2 files changed, 3 insertions(+), 6 deletions(-)
drivers/bluetooth/hci_ldisc.c +3 -5
@@ -470,8 +470,6 @@ static int hci_uart_tty_open(struct tty_struct *tty)INIT_WORK (&hu->init_ready, hci_uart_init_work);INIT_WORK (&hu->write_work, hci_uart_write_work); -spin_lock_init (&hu->rx_lock); -/* Flush any pending characters in the driver and line discipline. */ /* FIXME: why is this needed. Note don't use ldisc_ref here as the
@@ -569,14 +567,14 @@ static void hci_uart_tty_receive(struct tty_struct *tty, const u8 *data,if (!test_bit (HCI_UART_PROTO_SET , &hu->flags))return ; -spin_lock (&hu->rx_lock); +/* It does not need a lock here as it is already protected by a mutex in +* tty caller +*/ hu->proto->recv (hu, data, count);if (hu->hdev) hu->hdev->stat.byte_rx += count; -spin_unlock (&hu->rx_lock); -tty_unthrottle (tty); }
drivers/bluetooth/hci_uart.h +0 -1
@@ -85,7 +85,6 @@ struct hci_uart {struct sk_buff *tx_skb;unsigned long tx_state; -spinlock_t rx_lock;unsigned int init_speed;unsigned int oper_speed;
Signature · 1 line
1.9.1
PATCH v5 5/5 Bluetooth: hci_bcm: Add suspend/resume runtime PM functions
In reply to Frederic DanisAdds autosuspend runtime functionality to BCM UART driver. Autosuspend is enabled at end of bcm_setup.
bcm_device_lock is used for system sleep functions as they can be called at any time.bcm_device_lock is not held for runtime suspend functions as this is only enabled as long as platform device is opened.
--- drivers/bluetooth/hci_bcm.c | 60 ++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 56 insertions(+), 4 deletions(-)
drivers/bluetooth/hci_bcm.c +56 -4
@@ -33,6 +33,7 @@#include <linux/tty.h> #include <linux/interrupt.h> #include <linux/dmi.h> +#include <linux/pm_runtime.h> #include <net/bluetooth/bluetooth.h> #include <net/bluetooth/hci_core.h>
@@ -40,6 +41,8 @@#include "btbcm.h" #include "hci_uart.h" +#define BCM_AUTOSUSPEND_DELAY 5000/* default autosleep delay */ +struct bcm_device {struct list_head list;
@@ -160,6 +163,10 @@ static irqreturn_t bcm_host_wake(int irq, void *data)bt_dev_dbg (bdev,"Host wake IRQ" ); +pm_runtime_get (&bdev->pdev->dev); +pm_runtime_mark_last_busy (&bdev->pdev->dev); +pm_runtime_put_autosuspend (&bdev->pdev->dev); +return IRQ_HANDLED ; }
@@ -183,6 +190,12 @@ static int bcm_request_irq(struct bcm_data *bcm) goto unlock;device_init_wakeup (&bdev->pdev->dev, true); + +pm_runtime_set_autosuspend_delay (&bdev->pdev->dev, +BCM_AUTOSUSPEND_DELAY ); +pm_runtime_use_autosuspend (&bdev->pdev->dev); +pm_runtime_set_active (&bdev->pdev->dev); +pm_runtime_enable (&bdev->pdev->dev); } unlock:
@@ -198,7 +211,7 @@ static const struct bcm_set_sleep_mode default_sleep_params = { .bt_wake_active = 1, /* BT_WAKE active mode: 1 = high, 0 = low */ .host_wake_active = 0, /* HOST_WAKE active mode: 1 = high, 0 = low */ .allow_host_sleep = 1, /* Allow host sleep in SCO flag */ - .combine_modes = 0, /* Combine sleep and LPM flag */ + .combine_modes = 1, /* Combine sleep and LPM flag */ .tristate_control = 0, /* Allow tri-state control of UART tx flag *//* Irrelevant USB flags */ .usb_auto_sleep = 0,
@@ -284,6 +297,9 @@ static int bcm_close(struct hci_uart *hu)if (bcm_device_exists (bdev)) {bcm_gpio_set_power (bdev, false);#ifdef CONFIG_PM +pm_runtime_disable (&bdev->pdev->dev); +pm_runtime_set_suspended (&bdev->pdev->dev); +if (device_can_wakeup (&bdev->pdev->dev)) {devm_free_irq (&bdev->pdev->dev, bdev->irq, bdev);device_init_wakeup (&bdev->pdev->dev, false);
@@ -400,6 +416,15 @@ static int bcm_recv(struct hci_uart *hu, const void *data, int count)bt_dev_err (hu->hdev,"Frame reassembly failed (%d)" , err); bcm->rx_skb =NULL ;return err; + }else if (!bcm->rx_skb) { +/* Delay auto-suspend when receiving completed packet */ +mutex_lock (&bcm_device_lock); +if (bcm->dev &&bcm_device_exists (bcm->dev)) { +pm_runtime_get (&bcm->dev->pdev->dev); +pm_runtime_mark_last_busy (&bcm->dev->pdev->dev); +pm_runtime_put_autosuspend (&bcm->dev->pdev->dev); + } +mutex_unlock (&bcm_device_lock); }return count;
@@ -421,8 +446,27 @@ static int bcm_enqueue(struct hci_uart *hu, struct sk_buff *skb)static struct sk_buff *bcm_dequeue (struct hci_uart *hu) {struct bcm_data *bcm = hu->priv; +struct sk_buff *skb =NULL ; +struct bcm_device *bdev =NULL ; + +mutex_lock (&bcm_device_lock); + +if (bcm_device_exists (bcm->dev)) { + bdev = bcm->dev; +pm_runtime_get_sync (&bdev->pdev->dev); +/* Shall be resumed here */ + } + + skb =skb_dequeue (&bcm->txq); + +if (bdev) { +pm_runtime_mark_last_busy (&bdev->pdev->dev); +pm_runtime_put_autosuspend (&bdev->pdev->dev); + } -return skb_dequeue (&bcm->txq); +mutex_unlock (&bcm_device_lock); + +return skb; }#ifdef CONFIG_PM
@@ -490,7 +534,8 @@ static int bcm_suspend(struct device *dev)if (!bdev->hu) goto unlock; -bcm_suspend_device (dev); +if (pm_runtime_active (dev)) +bcm_suspend_device (dev);if (device_may_wakeup (&bdev->pdev->dev)) { error =enable_irq_wake (bdev->irq);
@@ -530,6 +575,10 @@ static int bcm_resume(struct device *dev) unlock:mutex_unlock (&bcm_device_lock); +pm_runtime_disable (dev); +pm_runtime_set_active (dev); +pm_runtime_enable (dev); + return 0; }#endif
@@ -750,7 +799,10 @@ MODULE_DEVICE_TABLE(acpi, bcm_acpi_match);#endif /* Platform suspend and resume callbacks */ -static SIMPLE_DEV_PM_OPS(bcm_pm_ops, bcm_suspend, bcm_resume); +static const struct dev_pm_ops bcm_pm_ops = { + SET_SYSTEM_SLEEP_PM_OPS(bcm_suspend, bcm_resume) +SET_RUNTIME_PM_OPS (bcm_suspend_device, bcm_resume_device,NULL ) +};static struct platform_driver bcm_driver = { .probe = bcm_probe,
Signature · 1 line
1.9.1
Marcel Holtmann
maintainerHi Fred,
Quoted 44 lines
Add wake-up capabilities by retrieveing interruption used by BCM device in ACPI table.
Add PM runtime support.
v4->v5:
- Addbt_warnto to unify Bluetooth traces inBT_WARN
- Change bcm_suspend_device/bcm_resume_device prototype to be usable directly inSET_RUNTIME_PM_OPS. This allows to remove bcm_runtime_suspend/bcm_runtime_resume which just call bcm_suspend_device/bcm_resume_device.
- Update commit message and add comment inhci_uart_tty_receiveto explain whyspin_lockis not requested
v3->v4:
- Add newBT_WARNandbt_dev_warnlogging macros
- Usebt_dev_warnfor T100 irq polarity fix
- Rename__bcm_suspendtobcm_suspend_device,
and__bcm_resumetobcm_resume_device
- Remove uselessrx_lockspinlock. It is only used inhci_uart_tty_receive()which is thereceive_bufldisc callback. This is already protected byflush_to_ldisc()indrivers/tty/tty_buffer.c.
This allows to use mutex inbcm_recv()and call pm_runtime_* directly
- Removepm_workWork queue which is now useless
- Delay auto-suspend only when reception of a packet is completed
v2->v3:
- UseDMI_EXACT_MATCHinstead ofDMI_MATCH
- Fix IRQ polarity for T100TA indriver_dataofdmi_system_idstruct
- Usedmi_first_match()instead ofdmi_check_system()
v1->v2:
- Split 1st patch between general wake-up capability and T100TA IRQ fix
- Replace multiple "if ... else if" by switch inbcm_resource()
- Move code to limit number of #ifdef
- Use DMI info to restrict IRQ to T100TA
- Split 2nd patch to prepare PM runtime support in separated patch
- Tested with and withoutCONFIG_PM_SLEEPandCONFIG_PM.Frederic Danis (5):
Bluetooth: AddBT_WARNandbt_dev_warnlogging macros
Bluetooth:hci_bcm: Fix IRQ polarity for T100
Bluetooth:hci_bcm: Prepare PM runtime support
Bluetooth: Remove uselessrx_lockspinlock
Bluetooth:hci_bcm: Add suspend/resume runtime PM functions
Quoted 6 lines
drivers/bluetooth/hci_bcm.c | 171 +++++++++++++++++++++++++++++++------- drivers/bluetooth/hci_ldisc.c | 8 +- drivers/bluetooth/hci_uart.h | 1 - include/net/bluetooth/bluetooth.h | 5 ++ net/bluetooth/lib.c | 16 ++++ 5 files changed, 164 insertions(+), 37 deletions(-)
all 5 patches have been applied to bluetooth-next tree.
Regards
Marcel