WD RE4 and APM

[JDurand]

Hello,

I have a big problem with RE4. The RE3 works fine, no problem. But the RE4 have many problem, like RAID 1 malfunction and raw read error.

When i disable APM on RE4, the error rate slowdown. This does not solve the problem entirely.

And for final user the big difference between RE3 and RE4 is : 1)  Cache of 64MB  ; 2) APM enable by default.

Or on wd website we can read : " All Western Digital Serial ATA hard drives come with PM2 (power management) disabled". (http://wdc.custhelp.com/app/answers/detail/a_id/958/~/how-to-enable-and-disable-pm2-(power-managemen...

Placing a jumper shunt on pins 3 and 4 does not disable APM. The disk cant start (plate dont turn).

Do you have information please ?

Thanks a lot.

[Awopero]

I guess you best shot here is contacting WD directly, if the drive does not perform properly for you, they will replace it.

[JDurand]

Thanks you Awopero.

It's not an isolated problem.

I have about 120 hosts with RE4 250Go (some RAID1 and others in standalone) and all have  the problem.

I have 20 disks new for testing and all have the problem.

I have received two RE4 500Go and they have the problem also.

[fzabkar]

The PM2 jumper enables PUIS (Power Up In Standby). It does not affect APM (Automatic Power Management). These are two different things.

PUIS is useful when there are several drives attached to the same power supply. If PUIS were disabled (default), then all drives would attempt to spin up at once, creating a heavy drain on the PSU. PUIS allows the OS to spin up each drive on demand, either via a read or write command, or via a special Set Features ATA command. PUIS can also be configured in firmware via an ATA command.

An alternative way to spin up a drive on demand is via pin 11 of the SATA power interface. This pin is often available to a RAID controller via a server backplane. However, in a desktop application this pin is usually grounded.

APM is a feature that enables a drive to autpark its heads after a preprogrammed period of inactivity. This timeout period is set in firmware. You may be able to define the timeout period, or disable APM altogether, by using WD's WDIDLE utility (wdidle3.exe).

The ATA commands are described in the following document:

Working Draft AT Attachment 8 - ATA/ATAPI Command Set (ATA8-ACS):
http://www.t13.org/documents/UploadedDocuments/docs2008/D1699r6a-ATA8-ACS.pdf

Here is the reverse engineered Linux version of WDIDLE:
http://idle3-tools.sourceforge.net/

Here is one source of WDIDLE:
http://www.synology.com/support/faq_images/enu/wdidle3.zip

[JDurand]

Thanks you very much fzabkar.

The linux idle3 is very interesting. On all my RE4, Timer are disabled. And i dont have problem with load cycle.

So the "Green" problem is another problem.

The ATA spec is very interesting. I think there is an ATA8 implementation problem in linux. But i dont know where. I'm reading ahci code, but i can't find anything. I was suspecting a buffer len problem with word 21 (ATA identity).

If the problem is in ahci driver ata_piix have the same...

[fzabkar]

Word #21 has been retired from the standard. It cannot report a buffer size greater than 32MB - 512 bytes.

See the following article for an explanation:
http://seagate.custkb.com/seagate/crm/selfservice/search.jsp?DocId=203631

[JDurand]

I know that Word 21 has been retired. But maybe some code part are using this parameter for setting something.

The question is, how RE4 can be uncompatible with linux (tested with kernel 2.6.32 and 3.0) while RE3 works fine.

RE4 :

ATA device, with non-removable media
        Model Number:       WDC WD2503ABYX-01WERA0
        Serial Number:      WD-WMAYP0721060
        Firmware Revision:  01.01S01
        Transport:          Serial, SATA 1.0a, SATA II Extensions, SATA Rev 2.5, SATA Rev 2.6

RE3 :

ATA device, with non-removable media
        Model Number:       WDC WD2502ABYS-02B7A0
        Serial Number:      WD-WCAT1H135940
        Firmware Revision:  02.03B03
        Transport:          Serial, SATA 1.0a, SATA II Extensions, SATA Rev 2.5

Command/Feature listed by hdparm :

jd@dt:~$ diff /tmp/RE3 /tmp/RE4
11a12
>            *    Advanced Power Management feature set
24a26
>            *    IDLE_IMMEDIATE with UNLOAD
31a34,35
>            *    Idle-Unload when NCQ is active
>            *    NCQ priority information 

Thanks for your help fzabkar.

[fzabkar]

Could "IDLE_IMMEDIATE with UNLOAD" be the answer? This feature is reported in bit 13 of word 84 of the Identify Device information block.

Section 7.19 of the ATA standard indicates that the above ATA command will cause a supported drive to park its heads in the landing zone, which means that the drive will spin down. Conversely, I suspect that a drive that doesn't support the same "IDLE_IMMEDIATE with UNLOAD" command may remain spinning in idle mode. Maybe that's the difference ???

BTW, I should have writtent that APM is Advanced Power Management, not Automatic Power Management. Sorry for my error.

[JDurand]

Hello fzabkar,

The news are not very good

After putting a lot of printk in kernel, i can't find any call to cmd IDLE IMMEDIATE.

Park function are not called.

I have played with kernel blacklist (ATA_HORKAGE_NONCQ ATA_HORKAGE_BROKEN_HPA ATA_HORKAGE_BROKEN_IPM). And this does not change anythings.

When  i run dd if=/dev/zero of=/tmp/bigfile the raw_read_error_rate is growing.

Forcing udma5 reduce the raw read error rate.

Any help from WD are welcome...

[JDurand · Accepted solution]

Hello all,

I have a very good news ! My problem is solve with the new firmware 01.01S02

After three months of cold war a nice guys at WD gave me the sesam : 01.01S02

The very very new harddrive come with this firmware, but not all.