On Fri, Feb 17, 2012 at 11:07 PM, Alan Stern [off-list ref] wrote:

On Fri, 17 Feb 2012, Zhang, Rui wrote:

quoted

quoted

Do you basically want
the ZPODD always to be suspended and resumed along with the ATA port,

No. ZPODD suspends itself, which put ZPODD to a SCSI low power state (NOT power off/D3_COLD).
And then it is the "Runtime PM core" that suspends ATA port after ZPODD being suspended.
And the .runtime_suspend callback for ATA port actually turns off the ZPODD power.

During resume, ATA port is resumed first because of the ACPI wakeup event.
But in fact, this wakeup event should be read as "ZPODD remote wakeup signal", thus runtime resume request is sent to ZPODD, done by Patch 3/6.

quoted

or should it be possible to suspend the ZPODD while the port remains
running?

Sure, but the power is still on at this time.

Then maybe you can use pm_runtime_no_callbacks() for the ZPODD device.
It's explained in Documentation/power/runtime_pm.txt, and I use it for
USB interfaces.

If pm_runtime_no_callbacks() is used, runtime PM sysfs attributes
won't be created.
Then how to disable ZPODD feature in userspace?

Currently, I use "control" file of scsi device to enable/disable
ZPODD, for example
echo auto > /sys/devices/pci0000:00/0000:00:1f.2/ata0/host1/target1:0:0/1:0:0:0/power/control
echo on > /sys/devices/pci0000:00/0000:00:1f.2/ata0/host1/target1:0:0/1:0:0:0/power/control

The idea is that the ZPODD will never receive any runtime PM callbacks
from the PM core.  Instead the ATA port callback routines will be
responsible for power management of the ZPODD device.

Does the ATA port callback also responsible to resume its child?

For example,
/sys/devices/pci0000:00/0000:00:1f.2/ata0/host1/target1:0:0/1:0:0:0/

ata0 is resumed.

Then who will be responsible to resume host1, target1:0:0 and 1:0:0:0?

Or do you mean that we don't need to resume these devices at all?
host1 and target1:0:0 are logical devices, but I think 1:0:0:0 is not.

Thanks,
Lin Ming