Thread (6 messages) 6 messages, 5 authors, 2015-10-09

RE: [PATCH] input:gpio-key: set IRQF_NO_SUSPEND for wake capable key

flat view

From: Li, Aubrey <hidden>
Date: 2015-10-09 02:47:45

On Friday, October 09, 2015 10:14 AM, Zheng, Qi wrote:
Hi,

"enable_irq_wake" is to enable/disable power-management wake-on of an
IRQ.
It is wake-on control not interrupt enable control.
In my opinion, to make sure the keys press working when system suspend, both
wake-on and interrupt should be enabled.
Besides,  for many platforms, the gpio irq chip don't implement the method
"enable_irq_wake" by setting "IRQCHIP_SKIP_SET_WAKE".
So, I think "IRQF_NO_SUSPEND" is necessary for this case.
We had a discussion on this before. Here is the thread:
https://lkml.org/lkml/2014/7/8/970

device_init_wakeup() is a general interface and supposed to be used to enable
wakeup for the wakeup sources. How gpio keys probe and call device_init_wakeup()
is a good reference for your case.

Thanks,
-Aubrey
BRs

Zheng Qi
PEG->IPG->EIG SH IO/LPSS team

-----Original Message-----
From: Dmitry Torokhov [mailto:dmitry.torokhov@gmail.com]
Sent: Saturday, October 3, 2015 1:42 AM
To: Linus Walleij <redacted>
Cc: Zha, Qipeng <qipeng.zha@intel.com>; Linux Input
[off-list ref]; Zheng, Qi [off-list ref]; Li, Aubrey
[off-list ref]
Subject: Re: [PATCH] input:gpio-key: set IRQF_NO_SUSPEND for wake capable
key

On Fri, Oct 02, 2015 at 03:26:27AM -0700, Linus Walleij wrote:
quoted
On Thu, Sep 17, 2015 at 9:06 AM, Qipeng Zha [off-list ref]
wrote:
quoted
quoted
When wakeup attribute is set, GPIO key is supposed to wake up the
system from system sleep state, So set IRQF_NO_SUSPEND flag to keep
IRQ enabled during suspend.

Signed-off-by: Qi Zheng <redacted>
Signed-off-by: Aubrey Li <redacted>
Signed-off-by: Qipeng Zha <qipeng.zha@intel.com>
Reviewed-by: Linus Walleij <redacted>

Should this not be tagged for stable?
I do not believe we need this at all. Otherwise every single driver for devices
that might be wakeup sources needs this flag set. Which would basically means
it is a noop and we should not suspend IRQ threads by default.

The dirver correctly calls enable_irq_wake() and platform should do whatever it
needs to make sure wakeup interrupt will be serviced.

Thanks.

--
Dmitry
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help