2014-11-12 06:41:38

by Kiran Kumar Raparthy

[permalink] [raw]
Subject: [RFC v6 3/3] usb: phy: hold wakeupsource when USB is enumerated in peripheral mode

From: Todd Poynor <[email protected]>

usb: phy: hold wakeupsource when USB is enumerated in peripheral mode

Some systems require a mechanism to prevent system to enter into suspend
state when USB is connected and enumerated in peripheral mode.

This patch provides an interface to hold a wakeupsource to prevent suspend.
PHY drivers can use this interface when USB is connected and enumerated in
peripheral mode.

A timed wakeupsource is temporarily held on USB disconnect events, to allow
the rest of the system to react to the USB disconnection (dropping host
sessions, updating charger status, etc.) prior to re-allowing suspend.

Cc: Felipe Balbi <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: Android Kernel Team <[email protected]>
Cc: John Stultz <[email protected]>
Cc: Sumit Semwal <[email protected]>
Cc: Arve Hj?nnev?g <[email protected]>
Cc: Benoit Goby <[email protected]>
Signed-off-by: Todd Poynor <[email protected]>
[kiran: Added context to commit message, squished build fixes
from Benoit Goby and Arve Hjønnevåg, changed wakelocks usage
to wakeupsource, merged Todd's refactoring logic and simplified
the structures and code and addressed community feedback]
Signed-off-by: Kiran Raparthy <[email protected]>
---
drivers/usb/phy/phy.c | 29 +++++++++++++++++++++++++++--
include/linux/usb/phy.h | 5 +++++
2 files changed, 32 insertions(+), 2 deletions(-)

diff --git a/drivers/usb/phy/phy.c b/drivers/usb/phy/phy.c
index 2b1039e..b5c5fd3 100644
--- a/drivers/usb/phy/phy.c
+++ b/drivers/usb/phy/phy.c
@@ -329,6 +329,7 @@ int usb_add_phy(struct usb_phy *x, enum usb_phy_type type)
int ret = 0;
unsigned long flags;
struct usb_phy *phy;
+ char wsource_name[40];

if (x->type != USB_PHY_TYPE_UNDEFINED) {
dev_err(x->dev, "not accepting initialized PHY %s\n", x->label);
@@ -351,6 +352,10 @@ int usb_add_phy(struct usb_phy *x, enum usb_phy_type type)
x->type = type;
list_add_tail(&x->head, &phy_list);

+ snprintf(wsource_name, sizeof(wsource_name), "vbus-%s",
+ dev_name(x->dev));
+ wakeup_source_init(&x->wsource, wsource_name);
+
out:
spin_unlock_irqrestore(&phy_lock, flags);
return ret;
@@ -402,6 +407,7 @@ void usb_remove_phy(struct usb_phy *x)

spin_lock_irqsave(&phy_lock, flags);
if (x) {
+ wakeup_source_trash(&x->wsource);
list_for_each_entry(phy_bind, &phy_bind_list, list)
if (phy_bind->phy == x)
phy_bind->phy = NULL;
@@ -446,13 +452,32 @@ int usb_bind_phy(const char *dev_name, u8 index,
EXPORT_SYMBOL_GPL(usb_bind_phy);

/**
- * usb_phy_set_event - set event to phy event
+ * usb_phy_set_event - set event to phy event and
+ * hold/temporarily hold wakeupsource
* @x: the phy returned by usb_get_phy();
*
- * This sets event to phy event
+ * This holds per-PHY wakeupsource/timed wakeupsource
*/
void usb_phy_set_event(struct usb_phy *x, unsigned long event)
{
+
x->last_event = event;
+
+ switch (event) {
+ case USB_EVENT_ENUMERATED:
+ __pm_stay_awake(&x->wsource);
+ break;
+
+ case USB_EVENT_NONE:
+ case USB_EVENT_ID:
+ case USB_EVENT_VBUS:
+ case USB_EVENT_CHARGER:
+ __pm_wakeup_event(&x->wsource,
+ msecs_to_jiffies(TEMPORARY_HOLD_TIME));
+ break;
+
+ default:
+ break;
+ }
}
EXPORT_SYMBOL_GPL(usb_phy_set_event);
diff --git a/include/linux/usb/phy.h b/include/linux/usb/phy.h
index 3c713ff..bc58fb8 100644
--- a/include/linux/usb/phy.h
+++ b/include/linux/usb/phy.h
@@ -12,6 +12,8 @@
#include <linux/notifier.h>
#include <linux/usb.h>

+#define TEMPORARY_HOLD_TIME 2000
+
enum usb_phy_interface {
USBPHY_INTERFACE_MODE_UNKNOWN,
USBPHY_INTERFACE_MODE_UTMI,
@@ -89,6 +91,9 @@ struct usb_phy {
/* for notification of usb_phy_events */
struct atomic_notifier_head notifier;

+ /* wakeup source */
+ struct wakeup_source wsource;
+
/* to pass extra port status to the root hub */
u16 port_status;
u16 port_change;
--
1.8.2.1


2014-11-18 15:55:16

by Felipe Balbi

[permalink] [raw]
Subject: Re: [RFC v6 3/3] usb: phy: hold wakeupsource when USB is enumerated in peripheral mode

Hi,

On Wed, Nov 12, 2014 at 12:11:19PM +0530, Kiran Kumar Raparthy wrote:
> From: Todd Poynor <[email protected]>
>
> usb: phy: hold wakeupsource when USB is enumerated in peripheral mode
>
> Some systems require a mechanism to prevent system to enter into suspend
> state when USB is connected and enumerated in peripheral mode.
>
> This patch provides an interface to hold a wakeupsource to prevent suspend.
> PHY drivers can use this interface when USB is connected and enumerated in
> peripheral mode.
>
> A timed wakeupsource is temporarily held on USB disconnect events, to allow
> the rest of the system to react to the USB disconnection (dropping host
> sessions, updating charger status, etc.) prior to re-allowing suspend.
>
> Cc: Felipe Balbi <[email protected]>
> Cc: Greg Kroah-Hartman <[email protected]>
> Cc: [email protected]
> Cc: [email protected]
> Cc: Android Kernel Team <[email protected]>
> Cc: John Stultz <[email protected]>
> Cc: Sumit Semwal <[email protected]>
> Cc: Arve Hj�nnev�g <[email protected]>
> Cc: Benoit Goby <[email protected]>
> Signed-off-by: Todd Poynor <[email protected]>
> [kiran: Added context to commit message, squished build fixes
> from Benoit Goby and Arve Hjønnevåg, changed wakelocks usage
> to wakeupsource, merged Todd's refactoring logic and simplified
> the structures and code and addressed community feedback]
> Signed-off-by: Kiran Raparthy <[email protected]>
> ---
> drivers/usb/phy/phy.c | 29 +++++++++++++++++++++++++++--
> include/linux/usb/phy.h | 5 +++++
> 2 files changed, 32 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/usb/phy/phy.c b/drivers/usb/phy/phy.c
> index 2b1039e..b5c5fd3 100644
> --- a/drivers/usb/phy/phy.c
> +++ b/drivers/usb/phy/phy.c
> @@ -329,6 +329,7 @@ int usb_add_phy(struct usb_phy *x, enum usb_phy_type type)
> int ret = 0;
> unsigned long flags;
> struct usb_phy *phy;
> + char wsource_name[40];
>
> if (x->type != USB_PHY_TYPE_UNDEFINED) {
> dev_err(x->dev, "not accepting initialized PHY %s\n", x->label);
> @@ -351,6 +352,10 @@ int usb_add_phy(struct usb_phy *x, enum usb_phy_type type)
> x->type = type;
> list_add_tail(&x->head, &phy_list);
>
> + snprintf(wsource_name, sizeof(wsource_name), "vbus-%s",
> + dev_name(x->dev));
> + wakeup_source_init(&x->wsource, wsource_name);

do you really need the device name on the wakeup source or would
usb-vbus suffice ? Hmm, I guess we could have several of these in the
same system, so maybe there's no other way.

> @@ -446,13 +452,32 @@ int usb_bind_phy(const char *dev_name, u8 index,
> EXPORT_SYMBOL_GPL(usb_bind_phy);
>
> /**
> - * usb_phy_set_event - set event to phy event
> + * usb_phy_set_event - set event to phy event and
> + * hold/temporarily hold wakeupsource
> * @x: the phy returned by usb_get_phy();
> *
> - * This sets event to phy event
> + * This holds per-PHY wakeupsource/timed wakeupsource
> */
> void usb_phy_set_event(struct usb_phy *x, unsigned long event)
> {
> +
> x->last_event = event;
> +
> + switch (event) {
> + case USB_EVENT_ENUMERATED:
> + __pm_stay_awake(&x->wsource);
> + break;
> +
> + case USB_EVENT_NONE:
> + case USB_EVENT_ID:
> + case USB_EVENT_VBUS:
> + case USB_EVENT_CHARGER:
> + __pm_wakeup_event(&x->wsource,
> + msecs_to_jiffies(TEMPORARY_HOLD_TIME));
> + break;
> +
> + default:
> + break;
> + }

much, much better. Completely hidden from PHY drivers.

> }
> EXPORT_SYMBOL_GPL(usb_phy_set_event);
> diff --git a/include/linux/usb/phy.h b/include/linux/usb/phy.h
> index 3c713ff..bc58fb8 100644
> --- a/include/linux/usb/phy.h
> +++ b/include/linux/usb/phy.h
> @@ -12,6 +12,8 @@
> #include <linux/notifier.h>
> #include <linux/usb.h>
>
> +#define TEMPORARY_HOLD_TIME 2000

minor nit:

Let's call this:

#define USB_PHY_DEFAULT_WAKEUP_SRC_TIMEOUT 2000 /* ms */

When resending, please drop that "RFC" from subject and make sure your
series is sent as a single thread, instead of three separate patches.

cheers

--
balbi


Attachments:
(No filename) (3.95 kB)
signature.asc (819.00 B)
Digital signature
Download all attachments

2014-11-18 16:51:22

by Kiran Kumar Raparthy

[permalink] [raw]
Subject: Re: [RFC v6 3/3] usb: phy: hold wakeupsource when USB is enumerated in peripheral mode

On 18 November 2014 21:25, Felipe Balbi <[email protected]> wrote:
> Hi,
>
> On Wed, Nov 12, 2014 at 12:11:19PM +0530, Kiran Kumar Raparthy wrote:
>> From: Todd Poynor <[email protected]>
>>
>> usb: phy: hold wakeupsource when USB is enumerated in peripheral mode
>>
>> Some systems require a mechanism to prevent system to enter into suspend
>> state when USB is connected and enumerated in peripheral mode.
>>
>> This patch provides an interface to hold a wakeupsource to prevent suspend.
>> PHY drivers can use this interface when USB is connected and enumerated in
>> peripheral mode.
>>
>> A timed wakeupsource is temporarily held on USB disconnect events, to allow
>> the rest of the system to react to the USB disconnection (dropping host
>> sessions, updating charger status, etc.) prior to re-allowing suspend.
>>
>> Cc: Felipe Balbi <[email protected]>
>> Cc: Greg Kroah-Hartman <[email protected]>
>> Cc: [email protected]
>> Cc: [email protected]
>> Cc: Android Kernel Team <[email protected]>
>> Cc: John Stultz <[email protected]>
>> Cc: Sumit Semwal <[email protected]>
>> Cc: Arve Hj�nnev�g <[email protected]>
>> Cc: Benoit Goby <[email protected]>
>> Signed-off-by: Todd Poynor <[email protected]>
>> [kiran: Added context to commit message, squished build fixes
>> from Benoit Goby and Arve Hjønnevåg, changed wakelocks usage
>> to wakeupsource, merged Todd's refactoring logic and simplified
>> the structures and code and addressed community feedback]
>> Signed-off-by: Kiran Raparthy <[email protected]>
>> ---
>> drivers/usb/phy/phy.c | 29 +++++++++++++++++++++++++++--
>> include/linux/usb/phy.h | 5 +++++
>> 2 files changed, 32 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/usb/phy/phy.c b/drivers/usb/phy/phy.c
>> index 2b1039e..b5c5fd3 100644
>> --- a/drivers/usb/phy/phy.c
>> +++ b/drivers/usb/phy/phy.c
>> @@ -329,6 +329,7 @@ int usb_add_phy(struct usb_phy *x, enum usb_phy_type type)
>> int ret = 0;
>> unsigned long flags;
>> struct usb_phy *phy;
>> + char wsource_name[40];
>>
>> if (x->type != USB_PHY_TYPE_UNDEFINED) {
>> dev_err(x->dev, "not accepting initialized PHY %s\n", x->label);
>> @@ -351,6 +352,10 @@ int usb_add_phy(struct usb_phy *x, enum usb_phy_type type)
>> x->type = type;
>> list_add_tail(&x->head, &phy_list);
>>
>> + snprintf(wsource_name, sizeof(wsource_name), "vbus-%s",
>> + dev_name(x->dev));
>> + wakeup_source_init(&x->wsource, wsource_name);
>
> do you really need the device name on the wakeup source or would
> usb-vbus suffice ? Hmm, I guess we could have several of these in the
> same system, so maybe there's no other way.
yeah true.
>
>> @@ -446,13 +452,32 @@ int usb_bind_phy(const char *dev_name, u8 index,
>> EXPORT_SYMBOL_GPL(usb_bind_phy);
>>
>> /**
>> - * usb_phy_set_event - set event to phy event
>> + * usb_phy_set_event - set event to phy event and
>> + * hold/temporarily hold wakeupsource
>> * @x: the phy returned by usb_get_phy();
>> *
>> - * This sets event to phy event
>> + * This holds per-PHY wakeupsource/timed wakeupsource
>> */
>> void usb_phy_set_event(struct usb_phy *x, unsigned long event)
>> {
>> +
>> x->last_event = event;
>> +
>> + switch (event) {
>> + case USB_EVENT_ENUMERATED:
>> + __pm_stay_awake(&x->wsource);
>> + break;
>> +
>> + case USB_EVENT_NONE:
>> + case USB_EVENT_ID:
>> + case USB_EVENT_VBUS:
>> + case USB_EVENT_CHARGER:
>> + __pm_wakeup_event(&x->wsource,
>> + msecs_to_jiffies(TEMPORARY_HOLD_TIME));
>> + break;
>> +
>> + default:
>> + break;
>> + }
>
> much, much better. Completely hidden from PHY drivers.
>
>> }
>> EXPORT_SYMBOL_GPL(usb_phy_set_event);
>> diff --git a/include/linux/usb/phy.h b/include/linux/usb/phy.h
>> index 3c713ff..bc58fb8 100644
>> --- a/include/linux/usb/phy.h
>> +++ b/include/linux/usb/phy.h
>> @@ -12,6 +12,8 @@
>> #include <linux/notifier.h>
>> #include <linux/usb.h>
>>
>> +#define TEMPORARY_HOLD_TIME 2000
>
> minor nit:
>
> Let's call this:
>
> #define USB_PHY_DEFAULT_WAKEUP_SRC_TIMEOUT 2000 /* ms */
>
> When resending, please drop that "RFC" from subject and make sure your
> series is sent as a single thread, instead of three separate patches.
Sure Felipe,Thanks for your time.
>
> cheers
>
> --
> balbi