
Commercial interlude. (No I don't work for Ubuntu.) It is cool that Ubuntu sent me 5 cds for free. And stickers as well. (Though my laptop running ubuntu still has some issues). So use Ubuntu if you aren't willing to run Gentoo)
Back to our (ir-)regularly scheduled program.
Turns out the memory error was at 422.7M (of a 512M module). Good, if it were in the first couple of megs things might be a little more complicated. But I'm happy the error is towards the end of my memory (chances are 50% that it will be). So I added mem=420 to my grub options and I'm happily running now (error free, albeit with 20% less memory). But at least now I can boot to a stable environment if I want to install the badram kernel patches and make my kernel ignore only the bad bits of ram. (Sure wish that patch was included in the gentoo-kernel....)
Funny weird. I ran memtest while installing Ubuntu's Dapper Drake a few
days ago and it was reporting vast amounts of my 512Megs of memory bad. I
wasn't sure if to fault the memtest of my mem chips, I went with the
former.
Interesting. Either the badmem or badram patches website had a similar
story. They had a bad piece of memory, ran memtest and somehow the memory
fixed itself.... Isn't hardware fun? Nothing like those
non-deterministic errors!
Generally speaking, gentoo-sources only carries a very lightweight list of
patches and stuff added on by Gentoo devs. Anything experimental usually
isn't in the tree. There's always mm-sources though, that might have it.
Two days?! Is that normal? I was recently troubleshooting a mem issue,
found this program and let it run for 3 hours, then decided the RAM was
probably good. Maybe I should crank at it again.
SMG-
It ran for 58 hours and found 20 errors. The same test failed on different
passes. If you are having hardware issues, I would probably run it
overnight...
This may also help:
Thanks for the advice: a Mythbuntu box I just built from old parts had
flashing caps/scroll and your solution targeted the problem (and its
solution) right away.