Hi, Colin
The irq in rtc-tps6586x is the alarm interrupt, I had tried to disable/enable it
as you said, but the system couldn't be resumed by wakealarm.
I think it's better to change the i2c-tegra.c

Thanks
Wei.


-----Original Message-----
From: ccross-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org [mailto:ccross-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org] On Behalf Of Colin Cross
Sent: Monday, March 28, 2011 2:34 PM
To: Wei Ni
Cc: khali-PUYAD+kWke1g9hUCZPvPmw@public.gmane.org; ben-linux-elnMNo+KYs3YtjvyW6yDsg@public.gmane.org; snanda-F7+t8E8rja9g9hUCZPvPmw@public.gmane.org; linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: Re: [PATCH] i2c: busses: i2c-tegra: Using suspend_noirq()/resume_noirq() for suspend/resume

On Sun, Mar 27, 2011 at 11:22 PM,  [off-list ref] wrote:

From: Wei Ni <redacted>

When use rtc-tps6586x wakealarm to suspend/resume,
it will show a lot error messages:
"tps6586x 4-0034: failed to read interrupt status
tps6586x 4-0034: failed reading from 0xb5"
After resume, the system will call tps6586x_irq() because of the alarm
interrupt. This irq handle will read tps6586x interrupt status (0xb5), but at
that time the i2c driver didn't resume yet, the system didn't call the
tegra_i2c_resume(), so the reading will be failed.
I2C driver needs to be suspended late and resumed early than other drivers.
So, I2C driver uses the suspend_noirq/resume_noirq callbacks of struct dev_pm_ops
for early/late power management system.

This sounds like a bug in rtc-tps6586x.  I think you need to call
disable_irq in the rtc-tps6586x suspend function and an enable_irq in
the resume function, which will delay the delivery of the wake irq
until the rtc driver (and its parent, the i2c driver) has been
resumed.