Comments on: Enabling BIOS options on a live server with no rebooting

By: Someone

Someone — Tue, 23 Aug 2011 21:46:46 +0000

Hi,

How do you calculate the value for PCI_HEADER_TYPE? Is it the same for every chipset or different?

I have the (7500) chipset and I keep getting "DCA disabled, enabling now." on each consecutive execution of dca_probe, so the value is not getting set.

I used the offset from the manual http://www-techdoc.intel.com/c..., see page 453.

thanks!

By: JY

JY — Fri, 17 Dec 2010 04:05:11 +0000

Joe, awesome hack! I'm currently struggling with the same issue on a HP DL380G6. I've already figured out the equivalent register to enable DCA on the 5520 chipset on the motherboard. However, it looks like I need to enable the ioatdma functionality as well. Quite inconveniently, HP did not provide an option in their bios. I've been flipping through the Intel 5520 chipset documentation to see if there's a way to enable it, but no dice so far. Any suggestions on where else I can look to adapt your hack?

By: Enabling DCA without accessing BIOS directly Drija

Enabling DCA without accessing BIOS directly Drija — Thu, 11 Nov 2010 23:48:05 +0000

[...] tried to enable Intel’s Direct Cache Access (DCA) using the method posted on this blog: Timetobleed? Are the effects reversible, i.e. if I reboot the server, will the registers be reset to the way [...]

By: Martin Beauchamp

Martin Beauchamp — Mon, 04 Oct 2010 23:44:01 +0000

Oh look, Ferdy figured it out:

http://www.mail-archive.com/ntop-misc@listgateway.unipi.it/msg01185.html

I guess s/he has a single physical processor or is crashing once in a while...

By: Martin Beauchamp

Martin Beauchamp — Thu, 30 Sep 2010 00:52:09 +0000

Thanks for the great post and utility. For my fellow Googlers, I figured I'd add the results of my research for trying to enable DCA on a Dell Poweredge R610. I found the following clue as to why this might not be a feature of Dell's BIOS.

From Intel's "Intel® 5520 and Intel® 5500 Chipset Specification Update April 2010"

39. DCA can block progress of other transactions in Dual-IOH, dual socket systems
Problem: Direct Cache Access transactions can block progress of other transactions in Dual-IOH,
dual socket systems.
Implication: The Dual-IOH, dual socket system will hang if the problem scenario occurs.
Workaround: On Dual-IOH, dual socket systems, disable DCA on both of the IOH parts.Note that the
workaround is not required for Dual-IOH, single socket systems.
Status: Plan fix.

Maybe this will pop up in the R610 BIOS after Intel addresses the underlying issue? If its resolved by Intel and Dell doesn't enable DCA control in the BIOS, I was unable to find the address of where to enable DCA with this tool. Maybe someone else can figure out where to look for the new IOH architecture of the 5520...

By: TNAPI | 随同个人

TNAPI | 随同个人 — Wed, 18 Aug 2010 04:21:25 +0000

[...] has DCA disabled and the BIOS do not allow me to enable it. What shall I do?A. Have a look at this [...]

By: Rayne

Rayne — Wed, 14 Apr 2010 09:25:55 +0000

Great article!

Since this has to be done each time the server reboots, does it mean that the effects of doing this is not irreversible? For example, if I find that something isn't working right after doing this, I can simply restart my server and things will go back to the way it was?

By: Useful kernel and driver performance tweaks for your Linux server at time to bleed by Joe Damato

Tue, 28 Jul 2009 10:20:32 +0000

[...] Intel’s I/OAT also includes a feature called Direct Cache Access (DCA). DCA allows a driver to warm a CPU cache. A few NICs support DCA, the most popular (to my knowledge) is the Intel 10GbE driver (ixgbe). Refer to your NIC driver documentation to see if your NIC supports DCA. To enable DCA, a switch in the BIOS must be flipped. Some vendors supply machines that support DCA, but don’t expose a switch for DCA. If that is the case, see my last blog post for how to enable DCA manually. [...]

By: ice799

ice799 — Tue, 07 Jul 2009 02:40:18 +0000

Yep – but being that pedantic isn't particularly useful. I wanted to avoid spending time explaining MSRs and how they are used to keep my blog post short and to the point. Perhaps in a future blog post I will go into more detail here.

I thought that the first line of my article pointed out that I was describing a way to toggle CPU and chipset registers.

Thanks for reading.

By: ice799

ice799 — Tue, 07 Jul 2009 02:29:59 +0000

Yep, the driver could be patched to do this, too.