I work for an OEM of a tablet PC, we want to do something simalar to IBM's
'Active Protection system', where they use an acceloromitor to detect when
the tablet is dropped and shutdown the hard drive. In looking through the
DDK documantation I'm at a loss as to where to start, I though about
writing a HD filter driver and when needed sending a shutdown packet to the
hard disk driver but then how do I keep the rest of the system from bringing
it back up too soon?

Thanks,

Do IBM drives or controllers have a proprietary command for this?
--PA

"ms news" <awinner@walkabout-comp.com> wrote in message news:#QkhIqfZFHA.3032@TK2MSFTNGP10.phx.gbl...

Uhmm, this seems like something that would best be done in hardware.
Windows is non-deterministic wrt latencies, thus there is no way to
guarantee that the HD would be shut down before the tablet hit the floor.

--
Bill McKenzie


"ms news" <awinner@walkabout-comp.com> wrote in message
news:%23QkhIqfZFHA.3032@TK2MSFTNGP10.phx.gbl...

You should be able to send a SCSIOP_START_STOP_UNIT to the disk device.

cdb.START_STOP.OperationCode = SCSIOP_START_STOP_UNIT;
cdb.START_STOP.Immediate = 1;
cdb.START_STOP.Start = 0;
cdb.START_STOP.LoadEject = 0;

But you should flush the file systems before spinning down the disk.

This posting is provided "AS IS" with no warranties, and confers no rights.

Which is going to suffer from 10 ms quantums. The problem is 32 feet per
second per second, and how long it takes a laptop to fall from an average
table top height (3 feet?). Do you REALLY have enough time to stop the HDD
from spinning by sending it any SCSIOP_ command when you consider that you
just burned a good percentage of your time determining the dumb schmuck just
knocked me off the table.

My best guess is that this is NOT done in the OS at all, but via hardware.

--
Gary G. Little

""Bart Bartel [MSFT]"" <bbartel@online.microsoft.com> wrote in message
news:FlnZ32iZFHA.3928@TK2MSFTNGXA01.phx.gbl...

Gary, you've got me interested in playing math-. Let's assume the laptop
is dropped vertically with initial velocity =0, we have:

/ T
|
Height = | v(t)*dt
| 0
/
where:
T is the moment the laptop hit the floor.
v(t)=g*t and g=9.8 in most part of the planet.

so we have:
/ T
|
Height = | 9.8*t*dt
| 0
/


according to Newton-Leibniz Formula, we have:

Height= F(T)-F(0) where F(t)= 0.5*9.8*t*t

or:
T=(2*Height/9.8)^0.5 <<<<< formula 1

We substitute the average height is 3 feet, or 0.9144 meters to formula 1,
we then get T=0.43199 sec. or 431.99 ms.

With that said, we concluded:
The poor laptop would have more than enough time to shutdown his HDD before
he hit the floor from a 3-feet high table top.

Not that bad, eh?


--
Calvin Guan Windows DDK MVP
Staff SW Engineer, NetXtreme MINIPORT
Enterprise Network Controller Engineering
Broadcom Corporation www.broadcom.com




"Gary G. Little" <glittle@mn.rr.com> wrote in message
newsuane.5026$lI2.3283@tornado.rdc-kc.rr.com...

Pretty much our conclusion too, besides doing it in hardware is really
messy, trying to flush the buffer, keep the OS happy when you yank away the
drive,......

The only downside to the time is we don't have enough time to play a .wav
file, 'HELP me I'm falling':})

Given that this in in a filter driver as low in the stack as I can get it, I
have a few other questions:

Does the START_STOP flush the buffers, or do I need to do that first?

Should I lock the queue, during the free fall till post impact, or just self
queue any IRP that comes in, and then pass it along after I give the all
clear and turn back on the drive?

"Calvin Guan" <hguan@nospam.broadcom.com> wrote in message
news:Olwz39mZFHA.1456@TK2MSFTNGP15.phx.gbl...

Thanks,

This should get me started, why doesn't MS put this kind of stuff in the
docs, I should know better by now....


"Bart Bartel [MSFT]" <bbartel@online.microsoft.com> wrote in message
news:FlnZ32iZFHA.3928@TK2MSFTNGXA01.phx.gbl...

This gets back to my original point. You probably have control in a tablet
as to what other hardware gets to play, but in general you could have other
devices which could potentially rob all of your HD shutdown time.
Particularly if you had an aggressive video adapter and you were playing
Doom 3 when you dropped the machine You may not have 432ms. You may not
have any time, who knows?

My point is, there is no forced limitation in Windows as to how long a
driver spends in its ISR routine. Thus your ISR could get starved out by
other interrupts. This is unlikely in general, but a hardware solution
would make this completely deterministic. I would be pretty surprised if
IBM did this sort of thing in software. Not shocked mind you, just
surprised.

--
Bill McKenzie


"Calvin Guan" <hguan@nospam.broadcom.com> wrote in message
news:Olwz39mZFHA.1456@TK2MSFTNGP15.phx.gbl...

Don't most harddrives use some sort of flash for caching/buffering nowadays
anyway? Seems a flash could buffer in the interim or some such. The OS
wouldn't even have to know the media was spun down. Or would that require
too much flash? Just curious.

--
Bill McKenzie


"ms news" <awinner@walkabout-comp.com> wrote in message
news:unJOQTqZFHA.2128@TK2MSFTNGP15.phx.gbl...

Flash is not fast enough and can't handle the number of erase/write cycles.
Even a million erase/write cycles would be far from sufficient for a hard
drive. Some flash is probably present to handle bad sectors and other
parameters that change very infrequently. I would expect most hard drive
caches to still be static ram, though some could be dynamic, but that
requires more circuitry and logic on the PCB.

"Bill McKenzie" <bill.mckenzie@nospam.conexant.com> wrote in message
news:uHRURCtZFHA.2412@TK2MSFTNGP10.phx.gbl...

How about a name other than 'awinner' and 'ms news'? Why would Microsoft
want to give someone an easy lawsuit victory? The SCSI specifications are
copyrighted. All you have to do is buy them or find copies of the drafts.
My only question is how you can be working on mass storage without the
necessary specs. The ATAPI specs are also for sale. Likewise for almost
all of the specs for standard hardware.

"ms news" <awinner@walkabout-comp.com> wrote in message
news:OFIPb8qZFHA.1424@TK2MSFTNGP15.phx.gbl...

And asking how IBM does it is a non-sequetor. Those of us that know can't
say without viollating an NDA with IBM. I highly doubt that IBM has this
technique in public domain.

--
Gary G. Little

"David J. Craig" <SeniorDriversWriter@shogunyoshimuni.com.net> wrote in
message news:%23h$6$1vZFHA.3132@TK2MSFTNGP09.phx.gbl...

Having worked with NT through XP now for about 10 years ... I still
wouldn't. I would say that MOST of the time you MIGHT get the disc spun down
in time. But all it takes is one time with any kind of latency and or
synchronization delay and you are looking at a Toasted laptop. All it takes
is once. Besides ... how much time does it take for the laptop to KNOW it is
falling with sufficient velocity, issue an interrupt, schedule a DPC,
trigger a worker thread to get into PASSIVE_LEVEL, and finally build the IRP
to send to the HBA, assuming you can talk to the miniport without going
through the storage stack. Are you assuming the command is sent at the TOP
of the arc? I really doubt that. How long will it take the accelerometer to
determine an "Oh ma god I'm FALLLING" event?

I'll keep my money on hardware.

--
Gary G. Little

"Calvin Guan" <hguan@nospam.broadcom.com> wrote in message
news:Olwz39mZFHA.1456@TK2MSFTNGP15.phx.gbl...

I suspect it is in the 'public domain' as far as knowledge goes, but not
free. IBM is a patent machine and they know from their mainframe
experiences that as soon as you implement an idea in hardware (and now
software too), someone will reverse engineer it and use it for free unless
they have a patent on it. If someone really wants to know, I would search
the patent database and/or watch patents awarded to IBM during the next two
years. They have to get the patent application in before it is 'published'
or they loose the right to apply, OSIB (or so I believe, maybe wrongly).

"Gary G. Little" <glittle@mn.rr.com> wrote in message
news:aUrne.18072$k43.11520@tornado.rdc-kc.rr.com...

Such drives (with flash buffer) were demoed by Samsung just a few weeks ago.
Also flash-only drives were announced.

"Bill McKenzie" <bill.mckenzie@nospam.conexant.com> wrote in message
news:uHRURCtZFHA.2412@TK2MSFTNGP10.phx.gbl...

Add a disk filter driver (lower), which will be "kinda power policy owner"
and will request the device power IRP after getting a notification about the
acceleration.
The notification is sent by the accelerometer driver. Design your way of
sending such a notification - Ex callback object or such.

--
Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation
maxim@storagecraft.com
http://www.storagecraft.com

"ms news" <awinner@walkabout-comp.com> wrote in message
news:%23QkhIqfZFHA.3032@TK2MSFTNGP10.phx.gbl...

Have you noticed that IBM uses special kind of drives with this system.
The most important here is that the head parks immediately, not
waiting for the disk to spin down. The whole process must complete in ~ 300 ms.
What I am afraid of is that the OEM wants to sell a purely software solution,
not really feasible with "economy class" noname disks.
--PA

"Maxim S. Shatskih" <maxim@storagecraft.com> wrote in message news:eLZij3zZFHA.3032@TK2MSFTNGP10.phx.gbl...