[ale] Bad memory, need to work around it

[Michael Trausch]

Hello.

We've had a lot of power spikes, interruptions, and so forth here.  It
has taken a toll on my system, and memory is starting to go bad.
However, it's a very, very small range; like 4K, that has bad bits in
it. I'd like to work around it.

I was doing some looking up and (don't have the link handy) someone
out there said that adding "memmap=SIZE$START" would do the trick, and
that the system would add that to the e820 map as reserved and that'd
be that.  Not so simple, it seems, because when I do that I get an
"early panic" message from the kernel, and it refuses to boot.

memtest86 reports single-bit errors in the memory at 0x10f8a900
through 0x10f8a9fff (probably about 30 addresses in there).  All of
them are bad on the same bit; (contents & 00000100) is always equal to
00000100 on those bad bytes.

Is there some way that I can get the Linux kernel to disregard that
area of memory?  I don't want to go buy 4 GB of RAM just because 4K of
my current RAM is unusable.  If the problem worsens over time (as I
expect it will) then I'll replace it later.  But I don't want to
replace it until the pain of working around bad memory is worse then
spending the money on the RAM, since things are pretty tight at the
moment...

    - Mike