It's a "bit fade" problem. I ran memtest86 all night. No errors as I expected. 
I've tried several times to locate the problem with memtest86 without success. 
I was hoping perhaps it had become severe enough for memtest86 to catch.

The commercial version has a fade test, but the delay between write and read is 
a few seconds. I need a fade test delay of at least 24 hours.

If I can find a simple example of a program that loads and runs in real address 
mode on x86 I can write a fade test. I downloaded the memtest86 code, but have 
not looked at it yet.

New, larger DIMMs are becoming increasingly attractive. For $150 I can double 
the memory with new Samsung DIMMs. As a practical matter, that is the correct 
solution. But I have this perverse enthusiasm for the technically proper 
solution of identifying the bad DIMM by running a test program.

While it's annoying that X11 is not starting up properly because of some issue 
with SMF I don't yet understand, it's very impressive to have recovered the 
system from corrupted zfs pools with no loss of data. The corruption was 
apparently just in memory.

Reg



  
_______________________________________________
openindiana-discuss mailing list
[email protected]
https://openindiana.org/mailman/listinfo/openindiana-discuss

Reply via email to