memtest86 and kernel badram

One of my 256MB modules has a bad region of memory -- from around 190MB to 220MB I have a intermittent hot bit showing up according to MemTest-86 v3.2.

I'll replace it eventually, but I thought I could buy a little time *and* learn something by utilizing the badram= option on the kernel boot.

But, memtest is giving me strange results when I select the "badram=" output.

The biggest confusion (for me) is memtest will often generate a badram list viz:

badram=0x0c00f5cc,0x0c00f5cc,0x0cc175c,0x0cc175c,....

i.e., pairs of identical hex values. I believe the 2nd hex value of any pair is (vaguely documented as) some sort of bit mask to denote the size of the region with the error at the address in the first hex value. Can anybody explain to me what I'm seeing here?

A second confusion is that, when the badram list starts getting 4 or 5 pairs of hex digits long, duplicate (or, triplicate) pairs start to appear. I.e.

badram=....,0x0cb975cc,0x0fffffffc,......,0x0cb975cc,0x0fffffffc,...

I think I can figure out how to get past that problem. But, these two anomalies leave me with an uneasy feeling about memtest and it's badram= output.

I believe my hot bit problem lies in (across) the 16MB region of 0x0c000000 to 0x0d000000. To simply eliminate that entire region, wouldn't I use a badram= with the following hex pairs?

badram=0x0c000000,0xff000000

Or, am I way wrong in my understanding of the kernel badram= parm.? (Which, BTW, is documented where? And, I don't mean documented for the kernel programmers. Where's the documentation for the folks that are suppose to actually use it?)

Tnx Jonesy -- Marvin L Jones jonz W3DHJ linux