Hello

I'm trying to set up EDAC to report on ECC memory and it's not working. AFAIK the components are all suitable but I'm a hardware dilettante and this is the first time I've built a kernel.

Here are the components.

• Motherboard: Gigabyte GA-MA770-DS3 Rev 1.0
• CPU: Athlon x2 Dual Core BE-2300
• Memory: Kingston KVR800D2E5K2/2G
• Kernel: 2.6.29.6. Slackware 13.0 32-bit huge-smp-2.6.29.6, rebuilt with
  CONFIG_EDAC=y
  CONFIG_EDAC_MM_EDAC=y
  CONFIG_EDAC_AMD76X=y
• User-space utilities: EDAC utils 0.16

The motherboard documentation says "(Note 3) Use of a CPU that supports ECC is required if you wish to install ECC memory". AFAIK the Athlon X2 range of CPUs do support ECC but ... ?

The memory description is "4GB 800MHz DDR2 ECC CL5 DIMM (Kit of 2)".

The new kernel does have EDAC support as evidenced by So far, so good but the user-space utility doesn't find the driversWhat to do?

Best

Charles

Gentoo states it as mask which means it is not stable enough. If you insist of using it, you should make sure sysfsutils and dmidecode are installed. Since it is a user space program, make sure the kernel has support for user space programs.

You do not need EDAC to make ECC memory function. It happens behind the scenes.

Thanks Electro

The information about EDAC not being stable enough is illuminating. I'm aiming to improve system availability, not decrease it! I understand that parity errors should be automatically corrected but was hoping that EDAC could be used to report on error corrections, giving advance warning of a degrading chip if corrections were being done in the same locality.

Given that EDAC is not stable enough, do you happen to know if any other reporting tools are available?

No I do not know anything else. The EDAC is new to me. Also Linux-ECC project is new. IMHO, it is for those people that are perfectionist that want to know if their computer is working perfectly.

No offense if you are paranoid of having a perfect computer that handles data perfectly, then you should not use computers. Also the word perfect is not actually a good word to use in the computer industry or any engineering field. A good word to use is doing the best you can to minimize errors. In any engineering field there is always errors. There is an old saying "there is always room for error."

You already did the best you can by buying ECC memory. Make sure you stay within temperature and voltage limits or else your setup will be unstable and unreliable.

If only!

I know only too well that computers do not and cannot work perfectly; if nothing else, extra-terrestrial gamma radiation causes all semi-conductor devices to malfunction at random times. More manageably, devices degrade with time and their systematic error-rates increase.

Disk monitoring, once only for top-end servers, is becoming commoner in workstations, at best giving us early warning of disk failure and allowing us to replace at a convenient time before catastrophic failure interrupts day-to-day work.

The same is happening with memory (and PCI devices, ...). EDAC can deliver similar warning of impending failure on workstations as has been available on high-end servers. Apparently EDAC is not mature yet but something like it is likely to become common and in a few years hardware monitoring will be taken for granted.