[v5,0/6] irqchip: qcom: pdc: Introduce irq_set_wake call

Message

Maulik Shah Aug. 22, 2020, 4:16 p.m. UTC

Changes in v5:
- Update commit subject in v4 patch 1
- Add more details to commit message in v4 patch 2
- Add change to enable wake irqs during suspend using new flag in irqchip
- Use this in PDC and qcom pinctrl driver to enable wakeirqs on suspend
- Make for loop more readable and add more details in commit in v4 patch 7

Changes in v4:
- Drop "Remove irq_disable callback from msmgpio irqchip" patch from v3
- Introduce irq_suspend_one() and irq_resume_one() callbacks
- Use the new callbacks to unmask wake interrupts during suspend
- Reset only pdc interrupts that are mapped in DTSI

Changes in v3:
- Drop gpiolib change (v2 patch 1) since its already in linux-next
- Add Acked-by Linus Walleij for v2 patch 2 and v2 patch 3.
- Address Stephen's comment to on v2 patch 3
- Address Stephen's comment to change variable to static on v2 patch 4.
- Add a new change to use return value from .irq_set_wake callback
- Add a new change to reset PDC irq enable bank during init time

Changes in v2:
- Fix compiler error on gpiolib patch

This series adds support to lazy disable pdc interrupt.

Some drivers using gpio interrupts want to configure gpio for wakeup using
enable_irq_wake() but during suspend entry disables irq and expects system
to resume when interrupt occurs. In the driver resume call interrupt is
re-enabled and removes wakeup capability using disable_irq_wake() one such
example is cros ec driver.

With [1] in documentation saying "An irq can be disabled with disable_irq()
and still wake the system as long as the irq has wake enabled".

The PDC IRQs are currently "unlazy disabled" (disable here means that it
will be masked in PDC & GIC HW GICD_ISENABLER, the moment driver invokes
disable_irq()) such IRQs can not wakeup from low power modes like suspend
to RAM since the driver chosen to disable this.

During suspend entry, no one re-enable/unmask in HW, even if its marked for
wakeup.

One solutions thought to address this problem was...During suspend entry at
last point, irq chip driver re-enable/unmask IRQs in HW that are marked for
wakeup. This was attemped in [2].

This series adds alternate solution to [2] by "lazy disable" IRQs in HW.
The genirq takes care of lazy disable in case if irqchip did not implement
irq_disable callback. Below is high level steps on how this works out..

a. During driver's disable_irq() call, IRQ will be marked disabled in SW
b. IRQ will still be enabled(read unmasked in HW)
c. The device then enters low power mode like suspend to RAM
d. The HW detects unmasked IRQs and wakesup the CPU
e. During resume after local_irq_enable() CPU goes to handle the wake IRQ
f. Generic handler comes to know that IRQ is disabled in SW
g. Generic handler marks IRQ as pending and now invokes mask callback
h. IRQ gets disabled/masked in HW now
i. When driver invokes enable_irq() the SW pending IRQ leads IRQ's handler
j. enable_irq() will again enable/unmask in HW

[1] https://www.spinics.net/lists/kernel/msg3398294.html
[2] https://patchwork.kernel.org/patch/11466021/

Maulik Shah (6):
  pinctrl: qcom: Set IRQCHIP_SET_TYPE_MASKED and IRQCHIP_MASK_ON_SUSPEND
    flags
  pinctrl: qcom: Use return value from irq_set_wake() call
  genirq/PM: Introduce IRQCHIP_ENABLE_WAKEUP_ON_SUSPEND flag
  pinctrl: qcom: Set IRQCHIP_ENABLE_WAKEUP_ON_SUSPEND flag
  irqchip: qcom-pdc: Set IRQCHIP_ENABLE_WAKEUP_ON_SUSPEND flag
  irqchip: qcom-pdc: Reset PDC interrupts during init

 drivers/irqchip/qcom-pdc.c         | 14 +++++++++++--
 drivers/pinctrl/qcom/pinctrl-msm.c | 11 +++++-----
 include/linux/irq.h                | 41 ++++++++++++++++++++------------------
 kernel/irq/debugfs.c               |  1 +
 kernel/irq/pm.c                    |  7 ++++++-
 5 files changed, 47 insertions(+), 27 deletions(-)

Comments

Linus Walleij Aug. 27, 2020, 10:48 p.m. UTC | #1

On Sat, Aug 22, 2020 at 6:17 PM Maulik Shah <[e-mail]> wrote:

> Changes in v5:

The series:
Acked-by: Linus Walleij <[e-mail]>

Feel free to merge this through the irqchip tree,

Yours,
Linus Walleij