Re: [patch 2a/3] Expose Power Management Policy option to users

[Kristen Carlson Accardi]

On Wed, 13 Jun 2007 08:26:48 -0700
James Bottomley <James.Bottomley@xxxxxxxxxxxx> wrote:

> On Tue, 2007-06-12 at 10:46 -0700, Kristen Carlson Accardi wrote:
> > Expose Power Management Policy option to users
> > 
> > This patch will modify the scsi subsystem to allow
> > users to set a power management policy for the link.
> > 
> > The scsi subsystem will create a new sysfs file for each
> > host in /sys/class/scsi_host called "link_power_management_policy".
> > This file can have 3 possible values:
> 
> I'm afraid the host isn't really the right place to put the link power
> management policy (assuming you want to manage the individual links
> separately) because there isn't a one to one correspondence between
> links and hosts.
> 
> To take the model I understand: SAS; the links are managed at the phy
> level, so the power policy should be set there and thus should probably
> be a property of the phy object, which doesn't even exist in the SCSI
> model, it only exists in the transport class.  It strikes me that even
> for ATA, the same thing is probably true.
> 
> Now, I can see that the power management models of all the transports
> might share some similarities (particularly at this three stage granular
> level); if so, it might make sense to export helpers from the mid-layer
> for the transport classes to use for this.

Ok - sorry for my ignorance about SCSI - but my sources (i.e. Arjan) tell 
me that the problem is that Link in ATA land means something different than 
Link in SCSI land, and that what I really need to do is leave this code under
the Host class, but rename it to something that more accurately reflects
what it means under SCSI.  So, is the word "segment" more appropriate?
Should I rename the file to "segment_power_management_policy"?

Thanks,
kristen


> 
> > Value		Meaning
> > -------------------------------------------------------------------
> > min_power	User wishes the link to conserve power as much as
> > 		possible, even at the cost of some performance
> > 
> > max_performance User wants priority to be on performance, not power
> > 		savings
> > 
> > medium_power	User wants power savings, with less performance cost
> > 		than min_power (but less power savings as well).
> 
> These seem like nicely sane and generic values.
> 
> James
> 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/