power off disk drives while running

power off disk drives while running

[Marty Leisner]

In embedded system design, it may be useful to poweroff the disks (as opposed
to merely spinning them down).  We want to leave the system running while
the disk is powered down, and let the disk powerup when it needs to be 
spun up.

While the "power off mechanism" would be platform dependent, is there a
generic path to announce "prepare for power going away"?


marty

Re: power off disk drives while running

[Jan Engelhardt]

On Aug 18 2007 12:08, Marty Leisner wrote:
>
>In embedded system design, it may be useful to poweroff the disks (as opposed
>to merely spinning them down).  We want to leave the system running while
>the disk is powered down, and let the disk powerup when it needs to be 
>spun up.

That means you also have to power it on...

>While the "power off mechanism" would be platform dependent, is there a
>generic path to announce "prepare for power going away"?

I do not see why that would be needed from a software point of view. Just make
sure that the disk does not needlessy emergency-park when pulling power. When
someone wants to write to disk, the request goes to the device driver, which
hands it to the controller, which hands it to the disk. And your controller
should be able to handle it (e.g. wait until reconnect) when there is a request
for a disk that is powered off.


	Jan
-- 

Re: power off disk drives while running

[Brennan Ashton]

On 8/18/07, Jan Engelhardt <jengelh@computergmbh.de> wrote:
>
> On Aug 18 2007 12:08, Marty Leisner wrote:
> >
> >In embedded system design, it may be useful to poweroff the disks (as opposed
> >to merely spinning them down).  We want to leave the system running while
> >the disk is powered down, and let the disk powerup when it needs to be
> >spun up.
>
> That means you also have to power it on...
>
> >While the "power off mechanism" would be platform dependent, is there a
> >generic path to announce "prepare for power going away"?
>
> I do not see why that would be needed from a software point of view. Just make
> sure that the disk does not needlessy emergency-park when pulling power. When
> someone wants to write to disk, the request goes to the device driver, which
> hands it to the controller, which hands it to the disk. And your controller
> should be able to handle it (e.g. wait until reconnect) when there is a request
> for a disk that is powered off.
>
>
>         Jan
> --
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

I see this a a very important feature in the embedded system relm, I
have worked on two projects that required extreme power management,
and massive data storage.  The ability to fully turn off a drive while
the system is running is key. It seems like this should be able to be
done from a kernel point of view rather than extra hardware. Although
if is not in the IDE/SATA spec then extra hardware would be necessary.
-- 
Brennan Ashton
Bellingham, Washington

"The box said, 'Requires Windows 98 or better'. So I installed Linux"

Re: power off disk drives while running

[Robert Hancock]

Brennan Ashton wrote:
> On 8/18/07, Jan Engelhardt <jengelh@computergmbh.de> wrote:
>> On Aug 18 2007 12:08, Marty Leisner wrote:
>>> In embedded system design, it may be useful to poweroff the disks (as opposed
>>> to merely spinning them down).  We want to leave the system running while
>>> the disk is powered down, and let the disk powerup when it needs to be
>>> spun up.
>> That means you also have to power it on...
>>
>>> While the "power off mechanism" would be platform dependent, is there a
>>> generic path to announce "prepare for power going away"?
>> I do not see why that would be needed from a software point of view. Just make
>> sure that the disk does not needlessy emergency-park when pulling power. When
>> someone wants to write to disk, the request goes to the device driver, which
>> hands it to the controller, which hands it to the disk. And your controller
>> should be able to handle it (e.g. wait until reconnect) when there is a request
>> for a disk that is powered off.
>>
>>
>>         Jan
>> --
>> -
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at  http://www.tux.org/lkml/
>>
> 
> I see this a a very important feature in the embedded system relm, I
> have worked on two projects that required extreme power management,
> and massive data storage.  The ability to fully turn off a drive while
> the system is running is key. It seems like this should be able to be
> done from a kernel point of view rather than extra hardware. Although
> if is not in the IDE/SATA spec then extra hardware would be necessary.

You can put a drive into sleep mode with ATA commands, that one requires 
a reset to take it out of that state (as opposed to standby which spins 
down but will spin up on any command that's issued afterwards). That's 
as close as it gets to fully powering off a drive through software.

-- 
Robert Hancock      Saskatoon, SK, Canada
To email, remove "nospam" from hancockr@nospamshaw.ca
Home Page: http://www.roberthancock.com/

Re: power off disk drives while running

[Jan Engelhardt]

On Aug 18 2007 14:22, Robert Hancock wrote:
>> I see this a a very important feature in the embedded system relm, I
>> have worked on two projects that required extreme power management,
>> and massive data storage.  The ability to fully turn off a drive while
>> the system is running is key. It seems like this should be able to be
>> done from a kernel point of view rather than extra hardware. Although
>> if is not in the IDE/SATA spec then extra hardware would be necessary.
>
> You can put a drive into sleep mode with ATA commands, that one requires a
> reset to take it out of that state (as opposed to standby which spins down but
> will spin up on any command that's issued afterwards). That's as close as it
> gets to fully powering off a drive through software.

An IDE reset bringing the disk up again -- that does not sound like
it is powered down. Power down for me means: as if the plug was pulled.


Well, you could also rewrite the standy ioctl to do this:

 - flush data
 - send spindown request
 - wait 1ms - 1s (give drive some time to park heads)
 - outportb(0x378, 0) - poweroff by setting LPT data line to 0
   (who knows? they might control the disk power!)

But you'd still have to fiddle with bringing it up again. That is, you have to
patch the block or device driver to outportb(0x378, 255) again when something
is supposed to spin up again.

Oh and of course you have to deal with the problem that all userspace apps may
hang because they are waiting for the disk.

Also consider that frequently spinning up/down is said to reduce lifetime.


	Jan
-- 

Re: power off disk drives while running

[Brennan Ashton]

On 8/18/07, Jan Engelhardt <jengelh@computergmbh.de> wrote:
>
> On Aug 18 2007 14:22, Robert Hancock wrote:
> >> I see this a a very important feature in the embedded system relm, I
> >> have worked on two projects that required extreme power management,
> >> and massive data storage.  The ability to fully turn off a drive while
> >> the system is running is key. It seems like this should be able to be
> >> done from a kernel point of view rather than extra hardware. Although
> >> if is not in the IDE/SATA spec then extra hardware would be necessary.
> >
> > You can put a drive into sleep mode with ATA commands, that one requires a
> > reset to take it out of that state (as opposed to standby which spins down but
> > will spin up on any command that's issued afterwards). That's as close as it
> > gets to fully powering off a drive through software.
>
> An IDE reset bringing the disk up again -- that does not sound like
> it is powered down. Power down for me means: as if the plug was pulled.
>
>
> Well, you could also rewrite the standy ioctl to do this:
>
>  - flush data
>  - send spindown request
>  - wait 1ms - 1s (give drive some time to park heads)
>  - outportb(0x378, 0) - poweroff by setting LPT data line to 0
>    (who knows? they might control the disk power!)
>
> But you'd still have to fiddle with bringing it up again. That is, you have to
> patch the block or device driver to outportb(0x378, 255) again when something
> is supposed to spin up again.
>
> Oh and of course you have to deal with the problem that all userspace apps may
> hang because they are waiting for the disk.
>
> Also consider that frequently spinning up/down is said to reduce lifetime.
>
>
>         Jan
> --
>
In my experience with embedded systems, usually the OS is residing in
flash with all of the apps. The hard disks are used a storage. Example
remote off grid high resolution time laps photography. the device
itself is simple OS, web server, and image grabber all reside in flash
32MB or so. Then HDs are used to store the images and turn off
completely at night and when disk is full.

-- 
Brennan Ashton
Bellingham, Washington

"The box said, 'Requires Windows 98 or better'. So I installed Linux"

Re: power off disk drives while running

[Phillip Susi]

Jan Engelhardt wrote:
> An IDE reset bringing the disk up again -- that does not sound like
> it is powered down. Power down for me means: as if the plug was pulled.

Reset means "power on, initialize power on state".  While shutdown in 
this way, the drive has shut down and is not drawing any power.  The 
reset is required to enable power flow into the drive again so it comes 
online.  Essentially you have a transistor disconnecting the power 
inside the drive, and the reset pulse flips that transistor.