Re: [PATCH v8 04/23] scsi: sd: Differentiate system and runtime start/stop management

[Phillip Susi <phill@thesusis.net>]

Damien Le Moal <dlemoal@kernel.org> writes:

> Your PMP is looking like type (1).

Correct, which is why the performance isn't good when I try to access
both disks at the same time.

>> When I enable runtime pm on the drives and the top level port, then try
>> to access the sleeping drive, I see this:
>
> How can you enable runtime PM on a port ? That does not work...

It works on *something*, I'm not not sure whether it is or is not the
port.  /sys/devices/pciblahblah/ata2/power/control can be enabled, and
then the port will be runtime suspended.  I'm not not sure what that is
compared to /sys/devices/pciblahblah/ata2/ata_port/ata2/power/control,
which can't be set to auto, and is also /sys/class/ata_port/ata2.  I'm
not sure why there are two different things that look like an "ata port".

> They are the first 2 devices connected to the PMP port that exist. These
> correspond to your sdd and sde scsi disks.

Right, but the PMP itself is not suspended.  I think that causes an
attempt to access the PMP while the root port that connects to the PMP
is suspended, so that access obviously fails.

> Are you running the latest code from Linus tree ? There is now a check on the
> power state of a device to avoid useless activate.

I hadn't pulled in some time.  Updating now.