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

[Damien Le Moal <dlemoal@kernel.org>]

On 11/7/23 22:27, Phillip Susi wrote:
> Damien Le Moal <dlemoal@kernel.org> writes:
> 
>> That is because the PM operations are defined for the *port*, not the *device*
>> (struct ata_device). With the missing PM ops for the device, PM core simply
>> sets the device as active.
> 
> The ata device appears to be entirely internal and not exposed as part
> of the sysfs device tree.  It is the generic block device that seems to
> be the one that actually implements runtime pm for the disk.
> 
> I think the problem is that the ata port is resumed first, then later
> the block device.  Thus, setting the block device to suspended in the
> port resume is later overruled by the system resume of the block
> device.  I'm thinking what is needed is for sd.c to query libata in its
> system resume callback and then set the runtime pm status depending on
> whether the drive is active or not.

Nope. The correct way to do this would be to define PM operations for the ata
device. However, currently, the scsi_device (scsi_disk) parent is the ata_port,
so to make sure that the PM status of the parent propagates to the child
correctly, we would need to have the scsi_device parent set to the ata_device.
As I said, correcting this is not simple and will involve a significant amount
of changes.

> By the way, I have noticed my system logs showing that the ata port that
> my dvd+rw is on is "Activating" twice, back to back.  Any idea why it
> would be trying to power up the optial drive twice?

Please send the exact messages you see in dmesg.

> 

-- 
Damien Le Moal
Western Digital Research