topic Re: Poison bit in Integrity Servers

Poison bit

Robert Cook_6 — Mon, 11 Dec 2006 12:24:32 GMT

My sx2000 went down yesterday for a poison bit set on a card in slot 3 of the pci card cage. This was part of the what.out file in the crash.0 directory. Can someone explain, as my boss put it, why a $200 pci card can bring down my million dollar server? Or could someone explain how the poison bit works.

Thanks,

Bob

MCA[0]:Platform Specific Non I/O error
MCA[0]:Platform Specific Data = 0xe000000000db8260
MCA[0]:Error Status val = 0x1a00
MCA[0]:Error Status type = ERR_POISONED A read was issued to data that has been
poisoned.
MCA/CMC[0]:handler = OS_MCA, sub_type = Generic
MCA/CMC[0]:00 00 00 00 00 00 20 05 e0 00 00 00 00 26 36 a0
MCA/CMC[0]:00 00 00 00 00 00 1a 00 00 00 00 00 00 00 00 00
MCA/CMC[0]:00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Re: Poison bit

Sameer_Nirmal — Tue, 12 Dec 2006 00:50:24 GMT

With mention of Poison Bit, it seems that you used a PCI-Express Card on the server with sx2000 chipset.

There was no support of PCI-Express until the HPUX 11i V2 September 2006 Release. The support for PCI-Express IO adapters was added in it's HWEnable11i bundle.

Refer information at
http://docs.hp.com/en/5991-5535/ch03s02.html

So if you don't have latest HWEnable11i bundle (Sept'06) installed, then it's not supported. Again then,it's not the PCI card who to blame, but to sx2000 chipset with support of PCI-Express.

The Poison bit ensures guaranteed error-freee packets. This bit provides an indicator to the end recipient that the data is corrupted, but still allows the data to be transferred. The notification is sent about it.

Refer the details of PCI-Express at
http://www.pcisig.com/specifications/pciexpress/

Re: Poison bit

Robert Cook_6 — Wed, 13 Dec 2006 08:52:00 GMT

Sameer,
Thank you for the information. The Pci card is a standard PCI card, not a PCI-E card. I found some information on a product called PCI Error Handling that looks like something I want to install. I am running Service Guard and have dual paths to my network and FC. The card that failed was an A9782 combo FC/NIC card. Do you have any knowledge of this PCI Error Handling product?

Thanks,

Bob

Re: Poison bit

Sameer_Nirmal — Wed, 13 Dec 2006 13:09:47 GMT

The PCI Error Handling product is available as PCIErrorHandling-00 bundle. It could be downloaded from
http://h20293.www2.hp.com/portal/swdepot/displayProductInfo.do?productNumber=PCIErrorHandling

However, take a look at the product note which is important for using this product in a specific environment.

As of June 2006, the iether and igelan Gigabit Ethernet drivers support PCI Error Handling when the PCIErrorHandling-00 bundle is installed on the system. Refer this link as well.
http://docs.hp.com/en/5991-5320/ch04s05.html

Re: Poison bit

Robert Cook_6 — Wed, 13 Dec 2006 13:26:22 GMT

Thanks,
I will install it on my development system and see how it works.

Bob