On Jan 6, 2004, at 06:06, Scott Ransom wrote:
If you are interested in trying my current 64bit kernel (2.6.1-rc1 + x86-64 group fixes), which is monolithic and has working gigabit and aic79xx, let me know and I can post mine somewhere.
If you have the patches and the .config file, could you post those somewhere?
[ BTW: To test for the bad memory, I just wrote a program that wrote 1gb of pseudo-random data, read it back, and compared, one 32-bit int at a time. When it hit bad memory, I got a MCE. ]Just curious: Why not use memtest?
Well, I tried running the modified memtest86 on this box, but it just rebooted the second it got above 4096GB. When I turned off ECC in the BIOS, it instead started spewing an unbelievable number of errors.
At the time, I thought memtest86 was just not able to handle >4GB. Now, OTOH, I bet it was actually the truth.
It found something like 40,000 errors on the first test alone. So I figured, there is no way someone could of actually shipped memory _THAT_ broken. (Especially after how well the other two servers we've gotten from them work.)
Well, later, I found out there is a way for someone to do that. Should of just asked here, I suppose; would of saved a fair bit of time, and much hair pulling.
Once I get replacement DIMMs, I'll be memtest86-ing the box for a week. At least. (With their previous 3/8's very bad failure [wonder if any of the remaining are bad, but ECC correctable --- hmmm, need to check] I expect to get at least one more bad DIMM...)