Re: SMART support for libata

Re: SMART support for libata

[Jeff Garzik]

Greg Stark wrote:
> I don't mean to be a pest, I'm just curious what the status is of SMART
> support for libata. It doesn't look like it's mentioned in any of the status
> updates. Did I miss it? Is it relatively low priority after the hardware
> support? Is anyone interested in working on it? What has to be done?

The code exists in libata-dev queue.  It's currently waiting on official 
  SCSI opcode assignments, and someone to add code to filter out SET 
FEATURES - XFER MODE commands.

	Jeff

Re: SMART support for libata

[Greg Stark]

FWIW here's a report of a bit of a problem with libata-dev and SMART support.
I'm not actually clear whether this is the fault of libata or the traditional
IDE drivers though.

I built 2.6.12rc4 with 2.6.11-bk6-libata-dev1.patch applied. 
(I had to fix up a couple things that didn't apply against 2.6.12)

I updated to 5.33 per someone's suggestion on the mailing list and SMART
support started working. Yay. At that point I noticed my old PATA drive was
getting really hot so I put it to sleep with "hdparm -Y".

Now whenever smartd probes that drive my system freezes for a few seconds and
I get this in my syslog:

Jun  8 12:49:36 stark kernel: hda: status timeout: status=0xd0 { Busy }
Jun  8 12:49:36 stark kernel: 
Jun  8 12:49:36 stark kernel: ide: failed opcode was: 0xe5

I'm fine with failing to get SMART info from a sleeping drive. I'm not clear
whether it's actually possible to get data back or not though it would be nice
to know how much sleeping is lowering the drive temperature. But freezing the
machine is unkind.

-- 
greg

Re: SMART support for libata

[Mark Lord]

Greg Stark wrote:
>
> getting really hot so I put it to sleep with "hdparm -Y".
> 
> Now whenever smartd probes that drive my system freezes for a few seconds and
> I get this in my syslog:
> 
> Jun  8 12:49:36 stark kernel: hda: status timeout: status=0xd0 { Busy }
> Jun  8 12:49:36 stark kernel: 
> Jun  8 12:49:36 stark kernel: ide: failed opcode was: 0xe5

That is normal and expected behaviour.
A "sleeping" drive never responds to commands
until woken with a reset.

You should be using "-y" (standby) instead of "-Y" (sleep).

Cheers

Re: SMART support for libata

[Greg Stark]

Mark Lord <liml@rtr.ca> writes:

> Greg Stark wrote:
> >
> > getting really hot so I put it to sleep with "hdparm -Y".
> > Now whenever smartd probes that drive my system freezes for a few seconds and
> > I get this in my syslog:
> > Jun  8 12:49:36 stark kernel: hda: status timeout: status=0xd0 { Busy }
> > Jun  8 12:49:36 stark kernel: Jun  8 12:49:36 stark kernel: ide: failed
> > opcode was: 0xe5
> 
> That is normal and expected behaviour.
> A "sleeping" drive never responds to commands
> until woken with a reset.

I'm fine with errors and SMART failing to get any data. 

It's the part about my entire computer freezing for 5-10s
that doesn't seem kosher to me.

> You should be using "-y" (standby) instead of "-Y" (sleep).

I'll try that. But that's not going to make it spin up when it gets a SMART
query is it?

-- 
greg

Re: SMART support for libata

[Mark Lord]

Greg Stark wrote:
..
>>You should be using "-y" (standby) instead of "-Y" (sleep).
> 
> I'll try that. But that's not going to make it spin up when it gets a SMART
> query is it?

Depends on what SMART items are being queried.

Actually, what you should *really* be using is "hdparm -S"
with a suitable timeout value (say, 30 or larger).

Cheers

Re: SMART support for libata

[Greg Stark]

Mark Lord <liml@rtr.ca> writes:

> Greg Stark wrote:
> ..
> >>You should be using "-y" (standby) instead of "-Y" (sleep).
> > I'll try that. But that's not going to make it spin up when it gets a SMART
> > query is it?
> 
> Depends on what SMART items are being queried.
> 
> Actually, what you should *really* be using is "hdparm -S"
> with a suitable timeout value (say, 30 or larger).

Not really since the drive will just spin up ever few seconds as bdflush (or
whatever it's called these days) dribbles out pages.

What I should *really* be using is the noflushd daemon. That's been on hold
since I found it didn't work with SATA drives. But I wonder if it would work
these days.

-- 
greg

Re: SMART support for libata

[Jens Axboe]

On Thu, Jun 09 2005, Greg Stark wrote:
> 
> Mark Lord <liml@rtr.ca> writes:
> 
> > Greg Stark wrote:
> > ..
> > >>You should be using "-y" (standby) instead of "-Y" (sleep).
> > > I'll try that. But that's not going to make it spin up when it gets a SMART
> > > query is it?
> > 
> > Depends on what SMART items are being queried.
> > 
> > Actually, what you should *really* be using is "hdparm -S"
> > with a suitable timeout value (say, 30 or larger).
> 
> Not really since the drive will just spin up ever few seconds as bdflush (or
> whatever it's called these days) dribbles out pages.
> 
> What I should *really* be using is the noflushd daemon. That's been on hold
> since I found it didn't work with SATA drives. But I wonder if it would work
> these days.

noflushd is ancient, have you tried playing with laptop mode?

-- 
Jens Axboe

Re: SMART support for libata

[Greg Stark]

Jens Axboe <axboe@suse.de> writes:

> > > Greg Stark wrote:
> > 
> > What I should *really* be using is the noflushd daemon. That's been on hold
> > since I found it didn't work with SATA drives. But I wonder if it would work
> > these days.
> 
> noflushd is ancient, have you tried playing with laptop mode?

Where do I find more about this?

-- 
greg

Re: SMART support for libata

[Adrian Bunk]

On Fri, Jun 10, 2005 at 11:35:44AM -0400, Greg Stark wrote:
> Jens Axboe <axboe@suse.de> writes:
> 
> > > > Greg Stark wrote:
> > > 
> > > What I should *really* be using is the noflushd daemon. That's been on hold
> > > since I found it didn't work with SATA drives. But I wonder if it would work
> > > these days.
> > 
> > noflushd is ancient, have you tried playing with laptop mode?
> 
> Where do I find more about this?

Documentation/laptop-mode.txt

> greg

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed