Andrew Morton akpm@linux-foundation.org writes:

On Fri, 31 May 2013 15:37:07 -0700 Kevin Hilman khilman@linaro.org wrote:

...

Currently, the RTC IRQ is never wakeup-enabled so is not capable of bringing the system out of suspend.

On OMAP platforms, we have gotten by without this because the TWL RTC is on an I2C-connected chip which is capable of waking up the OMAP via the IO ring when the OMAP is in low-power states.

However, if the OMAP suspends without hitting the low-power states (and the IO ring is not enabled), RTC wakeups will not work because the IRQ is not wakeup enabled.

To fix, ensure the RTC IRQ is wakeup enabled whenever the RTC alarm is set.

--- a/drivers/rtc/rtc-twl.c +++ b/drivers/rtc/rtc-twl.c @@ -213,12 +213,24 @@ static int mask_rtc_irq_bit(unsigned char bit) static int twl_rtc_alarm_irq_enable(struct device *dev, unsigned enabled) {

• struct platform_device *pdev = to_platform_device(dev);
• int irq = platform_get_irq(pdev, 0);
• static bool twl_rtc_wake_enabled; int ret;

• if (enabled)

• if (enabled) { ret = set_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);

• else

• if (device_can_wakeup(dev) && !twl_rtc_wake_enabled) {
  

• 	enable_irq_wake(irq);
  

• 	twl_rtc_wake_enabled = true;
  

• }
  

• } else { ret = mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);

• if (twl_rtc_wake_enabled) {
  

• 	disable_irq_wake(irq);
  

• 	twl_rtc_wake_enabled = false;
  

• }
  

• }

Why do we need this (slightly racy) logic with twl_rtc_wake_enabled? Other drivers don't do this.

I suspect the other drivers don't do this because they're making these calls in their suspend/resume callbacks and can make assumptions about the order and balance of the calls (e.g. the PM core will not call your resume callback if your suspend callback hasn't been called.) What's being protected against is imbalanced calling of the IRQ wake calls (e.g. calling disable_irq_wake() on an IRQ where enable_irq_wake() has not been called.)

In this patch, I'm using the rtc_alarm_irq_enable() function instead of the suspend/resume functions because of some limitations of parent device of this one (I described that at the end of a previous reponse to this thread[1].) Because of that, I cannot make the same assumptions about the ordering of enable/disable calls to this function.

Should we test device_may_wakeup() befre running disable_irq_wake()? Most drivers seem to do this, but it's all a bit foggy.

I suppose I could've added that check too for symmetry, but it would be redundant because the twl_rtc_wake_enabled flag can only be set when device_can_wakeup() == true.

Kevin

[1] http://marc.info/?l=linux-omap&m=137036801320665&w=2

Grygorii Strashko grygorii.strashko@ti.com writes:

Hi Kevin,

On 06/01/2013 01:37 AM, Kevin Hilman wrote:

...

Currently, the RTC IRQ is never wakeup-enabled so is not capable of bringing the system out of suspend.

On OMAP platforms, we have gotten by without this because the TWL RTC is on an I2C-connected chip which is capable of waking up the OMAP via the IO ring when the OMAP is in low-power states.

However, if the OMAP suspends without hitting the low-power states (and the IO ring is not enabled), RTC wakeups will not work because the IRQ is not wakeup enabled.

As I understand, IRQ wake up capabilities are set/clear simultaneously with IRQ unmasking/masking on OMAP4+ in omap-wakeupgen.c. So, it should work without this patch on OMAP4+.

It might work on OMAP4 for wakeup from suspend, but without properly declaring the IRQ as a wakeup source, it will not abort suspend if the RTC fires during the suspend process. To abort suspend, the IRQ must be declared as a wakeup IRQ.

But if TWL is used on non OMAP4+ platform then it is needed. (OMAP3: I haven't found the place where IRQ wakeup capabilities are configured, would be appreciate if you can point me on)

IRQ wakeup is a genirq feature that trickles into the irq_chip (in OMAP3 case, it's the twl4030 irq_chip.)

On OMAP3, as mentioned in the changelog, RTC wake has been working fine without this because we default to CORE retention, so wakeup happens via the IO ring. However, if you prevent retention during suspend, then this IRQ will not wake the system.

Kevin

...

To fix, ensure the RTC IRQ is wakeup enabled whenever the RTC alarm is set.

Cc: Alessandro Zummo a.zummo@towertech.it Cc: Andrew Morton akpm@linux-foundation.org Cc: Tony Lindgren tony@atomide.com Signed-off-by: Kevin Hilman khilman@linaro.org

---

drivers/rtc/rtc-twl.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/drivers/rtc/rtc-twl.c b/drivers/rtc/rtc-twl.c index 8751a52..bbda0fd 100644 --- a/drivers/rtc/rtc-twl.c +++ b/drivers/rtc/rtc-twl.c @@ -213,12 +213,24 @@ static int mask_rtc_irq_bit(unsigned char bit) static int twl_rtc_alarm_irq_enable(struct device *dev, unsigned enabled) {

• struct platform_device *pdev = to_platform_device(dev);
• int irq = platform_get_irq(pdev, 0);
• static bool twl_rtc_wake_enabled; int ret;
  • if (enabled)
• if (enabled) { ret = set_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);

• else

• if (device_can_wakeup(dev) && !twl_rtc_wake_enabled) {
  

• 	enable_irq_wake(irq);
  

• 	twl_rtc_wake_enabled = true;
  

• }
  

• } else { ret = mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);

• if (twl_rtc_wake_enabled) {
  

• 	disable_irq_wake(irq);
  

• 	twl_rtc_wake_enabled = false;
  

• }
  

• } return ret; }

twl-rtc has suspend/resume callbacks implemented, so I think it's the better place for this code and twl_rtc_wake_enabled can be dropped.

In theory, that might be the better place (and that's where I put these at first), but unfortunately, it doesn't work that way because the twl6030-irq core enables/diables the parent IRQ wake feature using PM notifiers (which was done to avoid potential lock recursion[1].)

During suspend, the notifier runs at suspend_prepare() time, which is well before the driver's ->suspend() method is called. The result is that the parents IRQ wakeup capabilies are never set.

Kevin

[1] commit ab2b9260df67e29d5bd69d989f2f84f8c2ed4238 Author: Todd Poynor toddpoynor@google.com Date: Tue Oct 4 11:52:29 2011 +0200

mfd: Fix twl6030 lockdep recursion warning on setting wake IRQs

LOCKDEP explicitly sets all irq_desc locks as a single lock-class, causing "possible recursive locking detected" when the TWL RTC driver calls through enable_irq_wake to twl6030_irq_set_wake, which recursively calls irq_set_irq_wake. Although the irq_desc and lock are different, LOCKDEP treats these as equivalent, presumably due to problems that can be incurred when locking more than one irq_desc, so best to avoid this.

Suspend/resume actions implemented as PM notifiers to avoid touch the TWL core for this.

Signed-off-by: Todd Poynor toddpoynor@google.com Acked-by: Santosh Shilimkar santosh.shilimkar@ti.com Signed-off-by: Samuel Ortiz sameo@linux.intel.com

Grygorii Strashko grygorii.strashko@ti.com writes:

On 06/04/2013 08:46 PM, Kevin Hilman wrote:

...

Grygorii Strashko grygorii.strashko@ti.com writes:

...

Hi Kevin,

On 06/01/2013 01:37 AM, Kevin Hilman wrote:

...

Currently, the RTC IRQ is never wakeup-enabled so is not capable of bringing the system out of suspend.

On OMAP platforms, we have gotten by without this because the TWL RTC is on an I2C-connected chip which is capable of waking up the OMAP via the IO ring when the OMAP is in low-power states.

However, if the OMAP suspends without hitting the low-power states (and the IO ring is not enabled), RTC wakeups will not work because the IRQ is not wakeup enabled.

As I understand, IRQ wake up capabilities are set/clear simultaneously with IRQ unmasking/masking on OMAP4+ in omap-wakeupgen.c. So, it should work without this patch on OMAP4+.

It might work on OMAP4 for wakeup from suspend, but without properly declaring the IRQ as a wakeup source, it will not abort suspend if the RTC fires during the suspend process. To abort suspend, the IRQ must be declared as a wakeup IRQ.

...

But if TWL is used on non OMAP4+ platform then it is needed. (OMAP3: I haven't found the place where IRQ wakeup capabilities are configured, would be appreciate if you can point me on)

IRQ wakeup is a genirq feature that trickles into the irq_chip (in OMAP3 case, it's the twl4030 irq_chip.)

On OMAP3, as mentioned in the changelog, RTC wake has been working fine without this because we default to CORE retention, so wakeup happens via the IO ring. However, if you prevent retention during suspend, then this IRQ will not wake the system.

Kevin

...

...

To fix, ensure the RTC IRQ is wakeup enabled whenever the RTC alarm is set.

Cc: Alessandro Zummo a.zummo@towertech.it Cc: Andrew Morton akpm@linux-foundation.org Cc: Tony Lindgren tony@atomide.com Signed-off-by: Kevin Hilman khilman@linaro.org

---

drivers/rtc/rtc-twl.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/drivers/rtc/rtc-twl.c b/drivers/rtc/rtc-twl.c index 8751a52..bbda0fd 100644 --- a/drivers/rtc/rtc-twl.c +++ b/drivers/rtc/rtc-twl.c @@ -213,12 +213,24 @@ static int mask_rtc_irq_bit(unsigned char bit) static int twl_rtc_alarm_irq_enable(struct device *dev, unsigned enabled) {

• struct platform_device *pdev = to_platform_device(dev);
• int irq = platform_get_irq(pdev, 0);
• static bool twl_rtc_wake_enabled; int ret;
  • if (enabled)
• if (enabled) { ret = set_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);

• else

• if (device_can_wakeup(dev) && !twl_rtc_wake_enabled) {
  

• 	enable_irq_wake(irq);
  

• 	twl_rtc_wake_enabled = true;
  

• }
  

• } else { ret = mask_rtc_irq_bit(BIT_RTC_INTERRUPTS_REG_IT_ALARM_M);

• if (twl_rtc_wake_enabled) {
  

• 	disable_irq_wake(irq);
  

• 	twl_rtc_wake_enabled = false;
  

• }
  

• } return ret; }

twl-rtc has suspend/resume callbacks implemented, so I think it's the better place for this code and twl_rtc_wake_enabled can be dropped.

In theory, that might be the better place (and that's where I put these at first), but unfortunately, it doesn't work that way because the twl6030-irq core enables/diables the parent IRQ wake feature using PM notifiers (which was done to avoid potential lock recursion[1].)

During suspend, the notifier runs at suspend_prepare() time, which is well before the driver's ->suspend() method is called. The result is that the parents IRQ wakeup capabilies are never set.

Sorry, forget about this patch - have no questions for this patch anymore.

Thanks.

Just FYI. It seems, The suspend will never be aborted on OMAP4 by SYSN_IRQ because of these two patches: 782baa2 mfd: Disable twl6030 IRQ during suspend 9c6079a genirq: Do not consider disabled wakeup irqs

You're right for the parent TWL IRQ, but the child interrupts (e.g. RTC) would still abort suspend if wakeup enabled.

Kevin