unexpected downspeed of AHCI controller on Intel Seaburg 5400

unexpected downspeed of AHCI controller on Intel Seaburg 5400

[starlight]

Have a relatively recent HP DL160 G5 that incorporates
the Seaburg 5400 chipset.  Appears the 5400 is a MCM
that includes a 631xESB/632xESB I/O controller.

Running untainted kernel 2.6.27.1.  Latest HP BIOS.

unexpected downspeed of AHCI controller on Intel Seaburg 5400

[starlight]

[prior incomplete message was sent accidently]

Have a relatively recent HP DL160 G5 incorporating the 
Seaburg 5400 chipset.  Appears the 5400 is a MCM including
a 631xESB/632xESB I/O controller.

Running kernel 2.6.27.1.  Latest HP BIOS.

Strangely the SATA controller seems to be reporting that
it can only run at 1.5 Gbps instead of 3 Gbps.

ahci 0000:00:1f.2: version 3.0
ahci 0000:00:1f.2: PCI INT B -> GSI 19 (level, low) -> IRQ 19
ahci 0000:00:1f.2: AHCI 0001.0100 32 slots 6 ports 1.5 Gbps 0x3f impl SATA mode
ahci 0000:00:1f.2: flags: 64bit ncq pm led slum part
ahci 0000:00:1f.2: setting latency timer to 64

Looked at the AHCI specification and Intel datasheets for both 
the 5400 and the 631xESB/632xESB.  The latter shows the register 
but does not include 3 Gpbs in the enumeration.  SATA-II support 
is stated in general for the device however.  The SATA links all 
are reported as running at 1.5 Gbps.

Definitely did pull the rate-limit jumpers off the Seagate
ES.2 drives.

details:

http://binnacle.cx/file/ahci_seaburg_downspeed_dmesg.txt
http://binnacle.cx/file/ahci_seaburg_downspeed_dmidecode.txt
http://binnacle.cx/file/ahci_seaburg_downspeed_hdparm-I.txt
http://binnacle.cx/file/ahci_seaburg_downspeed_lspci-vv.txt

Re: unexpected downspeed of AHCI controller on Intel Seaburg 5400

[Robert Hancock]

starlight@binnacle.cx wrote:
> [prior incomplete message was sent accidently]
> 
> Have a relatively recent HP DL160 G5 incorporating the 
> Seaburg 5400 chipset.  Appears the 5400 is a MCM including
> a 631xESB/632xESB I/O controller.
> 
> Running kernel 2.6.27.1.  Latest HP BIOS.
> 
> Strangely the SATA controller seems to be reporting that
> it can only run at 1.5 Gbps instead of 3 Gbps.
> 
> ahci 0000:00:1f.2: version 3.0
> ahci 0000:00:1f.2: PCI INT B -> GSI 19 (level, low) -> IRQ 19
> ahci 0000:00:1f.2: AHCI 0001.0100 32 slots 6 ports 1.5 Gbps 0x3f impl SATA mode

That comes right out of the controller's registers, so presumably it is 
telling the truth. That particular value is only used for display, 
though - the actual negotiated speed is determined from other registers. 
So if the controller says it can only do 1.5 and the drives all 
negotiated at 1.5, well, that's likely all it can do..

> ahci 0000:00:1f.2: flags: 64bit ncq pm led slum part
> ahci 0000:00:1f.2: setting latency timer to 64
> 
> Looked at the AHCI specification and Intel datasheets for both 
> the 5400 and the 631xESB/632xESB.  The latter shows the register 
> but does not include 3 Gpbs in the enumeration.  SATA-II support 
> is stated in general for the device however.  The SATA links all 
> are reported as running at 1.5 Gbps.

It doesn't necessarily mean anything that it supports SATA II, that's 
just a revision of the spec and it defines support for 3 Gbps, but 
doesn't require it.

> 
> Definitely did pull the rate-limit jumpers off the Seagate
> ES.2 drives.
> 
> details:
> 
> http://binnacle.cx/file/ahci_seaburg_downspeed_dmesg.txt
> http://binnacle.cx/file/ahci_seaburg_downspeed_dmidecode.txt
> http://binnacle.cx/file/ahci_seaburg_downspeed_hdparm-I.txt
> http://binnacle.cx/file/ahci_seaburg_downspeed_lspci-vv.txt
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ide" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: unexpected downspeed of AHCI controller on Intel Seaburg 5400

[starlight]

At 20:04 11/18/2008 -0600, Robert Hancock wrote:
>It doesn't necessarily mean anything that it supports SATA II, that's 
>just a revision of the spec and it defines support for 3 Gbps, but 
>doesn't require it.

Wasn't precise enough in my statement.  The Intel
datasheet says it supports 3.0 Gbps:

http://download.intel.com/design/chipsets/datashts/31308201.pdf

5.18 SATA Host Controller (D31:F2)

The Intel® 631xESB/632xESB I/O Controller Hub has an integrated 
SATA host controller that supports independent DMA operation on 
six ports and supports data transfer rates of up to 3.0 Gb/s 
(300 MB/s). . .

-----

The Seaburg 5400 is the absolute latest server chipset
with such fancy features as DCA.  Hard to believe that
Intel would not support 3 Gbps.

Is it possible for the BIOS to tell the chip to not negogiate 
above 1.5 Gbps?  I'm wondering if HP might have done this to 
prevent the bug below from happening.  If so it would be
nice for the driver to put it back to 3 Gbps.

http://www.intel.com/Assets/PDF/specupdate/313075.pdf

46. PxSERR errors are logged during BIOS initialization when 
running at 3Gbps on a small percentage of parts Problem: This 
failure signature has only been seen on a few Intel® 
631xESB/632xESB I/O Controller Hub parts and only happens when 
devices are running at 3Gbps. During BIOS boot up as the SATA 
controller is initialized, PxSERR errors are logged. Error bits 
1, 10, and 19 are set which are recovered data integrity error, 
protocol error, and 10b to 8b decode error. These bits are set 
in between BIOS postcode 423c and 503c. Some of these failing 
parts on certain platforms will hang at postcode 0x75 . Affects 
SATA ports.

Implication: May hang some operating systems\plaforms during 
boot up. Intel® 631xESB/632xESB I/O Controller Hub Specification 
Update 29

Errata
Workaround: NA.
Status: Fixed

Re: unexpected downspeed of AHCI controller on Intel Seaburg 5400

[Tejun Heo]

starlight@binnacle.cx wrote:
> Is it possible for the BIOS to tell the chip to not negogiate 
> above 1.5 Gbps?  I'm wondering if HP might have done this to 
> prevent the bug below from happening.  If so it would be
> nice for the driver to put it back to 3 Gbps.

The standard way to put that limit is to use SControl but it isn't doing
that, as far as ahci is concerned, it's free to negotiate any speed but
is negotiating 1.5Gbps.  Maybe they capped using some hidden register to
workaround the erratum.  At any rate, it doesn't really matter unless
you're gonna hang multiple devices to a single port using PMP.

Thanks.

-- 
tejun

Re: unexpected downspeed of AHCI controller on Intel Seaburg 5400

[starlight]

Thank you for explaining how it works.  Have resigned to living 
with 1.5G.  A reason exists to run the ES.2s at their full 
sequential 100 Mbytes/sec (800 Mbits/sec), and I had hoped that 
lowering the latency on the link might squeeze a bit more 
performance.  But 1.5 is good enough.



At 14:18 11/24/2008 +0900, Tejun Heo wrote:
>starlight@binnacle.cx wrote:
>> Is it possible for the BIOS to tell the chip to not negogiate 
>> above 1.5 Gbps?  I'm wondering if HP might have done this to 
>> prevent the bug below from happening.  If so it would be
>> nice for the driver to put it back to 3 Gbps.
>
>The standard way to put that limit is to use SControl but it isn't doing
>that, as far as ahci is concerned, it's free to negotiate any speed but
>is negotiating 1.5Gbps.  Maybe they capped using some hidden register to
>workaround the erratum.  At any rate, it doesn't really matter unless
>you're gonna hang multiple devices to a single port using PMP.
>
>Thanks.
>
>-- 
>tejun

Re: unexpected downspeed of AHCI controller on Intel Seaburg 5400

[Tejun Heo]

starlight@binnacle.cx wrote:
> Thank you for explaining how it works.  Have resigned to living 
> with 1.5G.  A reason exists to run the ES.2s at their full 
> sequential 100 Mbytes/sec (800 Mbits/sec), and I had hoped that 
> lowering the latency on the link might squeeze a bit more 
> performance.  But 1.5 is good enough.

Considering general disk operation speed and the parallel operation
achieved using NCQ, I don't think it will make any noticeable difference.

Thanks.

-- 
tejun