sata_promise / libata defaulting to UDMA/33

sata_promise / libata defaulting to UDMA/33

[Marc]

Hi. 

I'm having great difficulty solving a problem I've encountered. I have a
Promise Fastrak TX4000 card:

0000:00:08.0 RAID bus controller: Promise Technology, Inc. PDC20619 (FastTrak
TX4000) (rev 02)
        Subsystem: Promise Technology, Inc. PDC20619 (FastTrak TX4000)
        Flags: bus master, 66MHz, medium devsel, latency 96, IRQ 137

For some strange reason all drives are set to UDMA/33 mode when the module is
loaded and I haven't found a way of changing the UDMA mode (hdparm -X doesn't
work: HDIO_DRIVE_CMD(setxfermode) failed: Input/output error).

Extract from dmesg:

libata version 1.20 loaded.
sata_promise 0000:00:08.0: version 1.04
ata1: PATA max UDMA/133 cmd 0xF8ABC200 ctl 0xF8ABC238 bmdma 0x0 irq 137
ata2: PATA max UDMA/133 cmd 0xF8ABC280 ctl 0xF8ABC2B8 bmdma 0x0 irq 137
ata3: PATA max UDMA/133 cmd 0xF8ABC300 ctl 0xF8ABC338 bmdma 0x0 irq 137
ata4: PATA max UDMA/133 cmd 0xF8ABC380 ctl 0xF8ABC3B8 bmdma 0x0 irq 137
ata1: disabling port
scsi1 : sata_promise
ata2: dev 0 cfg 49:2f00 82:346b 83:7f01 84:4003 85:3469 86:3c01 87:4003 88:203f
ata2: dev 0 ATA-6, max UDMA/100, 312581808 sectors: LBA48
<b>ata2: dev 0 configured for UDMA/33</b>
scsi2 : sata_promise
ata3: dev 0 cfg 49:2f00 82:746b 83:7f01 84:4633 85:7469 86:3c01 87:4623 88:203f
ata3: dev 0 ATA-7, max UDMA/100, 312581808 sectors: LBA48
<b>ata3: dev 0 configured for UDMA/33</b>
scsi3 : sata_promise
ata4: dev 0 cfg 49:2f00 82:346b 83:7f01 84:4003 85:3469 86:3c01 87:4003 88:203f
ata4: dev 0 ATA-6, max UDMA/100, 312581808 sectors: LBA48
<b>ata4: dev 0 configured for UDMA/33</b>
scsi4 : sata_promise
  Vendor: ATA       Model: WDC WD1600JB-00E  Rev: 15.0
  Type:   Direct-Access                      ANSI SCSI revision: 05
  Vendor: ATA       Model: WDC WD1600JB-00R  Rev: 20.0
  Type:   Direct-Access                      ANSI SCSI revision: 05
  Vendor: ATA       Model: WDC WD1600JB-00E  Rev: 15.0
  Type:   Direct-Access                      ANSI SCSI revision: 05

I'm currently running kernel 2.6.17.3. In the controller's BIOS I've created 3
'stripe' arrays with 1 disk per array, the BIOS shows the 3 drives as U5 (UDMA
mode 5 - which is UDMA/100 right?).

I've searched high and low but haven't found an answer to this problem. Help
appreciated. Please - I'm getting dismal performance.

Regards,




--

Re: sata_promise / libata defaulting to UDMA/33

[Alan Cox]

Ar Gwe, 2006-08-04 am 00:31 +0800, ysgrifennodd Marc:
> Hi. 
> 
> I'm having great difficulty solving a problem I've encountered. I have a
> Promise Fastrak TX4000 card:

We knock things down to UDMA33 if they are on an unsuitable cable. The
rest of the displayed data looks correct.

> ata2: PATA max UDMA/133 cmd 0xF8ABC280 ctl 0xF8ABC2B8 bmdma 0x0 irq 137

Controller does 133

> ata2: dev 0 cfg 49:2f00 82:346b 83:7f01 84:4003 85:3469 86:3c01 87:4003 88:203f
> ata2: dev 0 ATA-6, max UDMA/100, 312581808 sectors: LBA48

Devices does UDMA 100
> <b>ata2: dev 0 configured for UDMA/33</b>

So this appears to be a cable confusion. One thing that was fixed in
recent patches was a bug where if the core code decided something was
PATA but the drive setup was in fact SATA it incorrectly clipped to
UDMA33.

> I'm currently running kernel 2.6.17.3. In the controller's BIOS I've created 3
> 'stripe' arrays with 1 disk per array, the BIOS shows the 3 drives as U5 (UDMA
> mode 5 - which is UDMA/100 right?).

Correct

> I've searched high and low but haven't found an answer to this problem. Help
> appreciated. Please - I'm getting dismal performance.

I think the needed patch is in 2.6.18-mm, not sure about 2.6.18-rc base.
Jeff Garzik would be able to verify what has it merged.

Re: sata_promise / libata defaulting to UDMA/33

[Marc]

On Thu, 03 Aug 2006 18:14:49 +0100, Alan Cox wrote
> Ar Gwe, 2006-08-04 am 00:31 +0800, ysgrifennodd Marc:
> > Hi. 
> > 
> > I'm having great difficulty solving a problem I've encountered. I have a
> > Promise Fastrak TX4000 card:
> 
> We knock things down to UDMA33 if they are on an unsuitable cable. 
> The rest of the displayed data looks correct.
> 
> > ata2: PATA max UDMA/133 cmd 0xF8ABC280 ctl 0xF8ABC2B8 bmdma 0x0 irq 137
> 
> Controller does 133
> 
> > ata2: dev 0 cfg 49:2f00 82:346b 83:7f01 84:4003 85:3469 86:3c01 87:4003
88:203f
> > ata2: dev 0 ATA-6, max UDMA/100, 312581808 sectors: LBA48
> 
> Devices does UDMA 100
> > <b>ata2: dev 0 configured for UDMA/33</b>
> 
> So this appears to be a cable confusion. One thing that was fixed in
> recent patches was a bug where if the core code decided something was
> PATA but the drive setup was in fact SATA it incorrectly clipped to
> UDMA33.
> 
> > I'm currently running kernel 2.6.17.3. In the controller's BIOS I've created 3
> > 'stripe' arrays with 1 disk per array, the BIOS shows the 3 drives as U5 (UDMA
> > mode 5 - which is UDMA/100 right?).
> 
> Correct
> 
> > I've searched high and low but haven't found an answer to this problem. Help
> > appreciated. Please - I'm getting dismal performance.
> 
> I think the needed patch is in 2.6.18-mm, not sure about 2.6.18-rc base.
> Jeff Garzik would be able to verify what has it merged.
> 
> -

Thank you. I've compiled and installed 2.6.18-rc3 and it now shows:

ata2: dev 0 configured for UDMA/100

Problem is the system locks up after a few minutes - not sure what's causing
that (no flashing keyboard LEDs so it's not a kernel panic). I'll try again
tomorrow - its very late here already!

Regards.

--