Hi Nathan et al,
> [Where mem is handed out]
Thanx! I [already] got it working for i386!
Scary: Linux didn't crash a single time during this development...
I will submit a patch as soon as I resolved some remaining trouble with the
maximum length of cmdline arguments.
> Of course, this is a system-by-system hand-patched fix. We are working to
> permit a user to specify sections of memory to exclude. (ie: smap=20M-24M:X)
My scheme is somewhat closer to hardware I think:
badram=adr0,mask0,adr1,mask1,...
where an address X is faulty if and only if (X & maskI) == (adrI & maskI)
for some I=0,1,...
This makes it simple (given experience with binary calculations) to capture
rows or columns from the RAM chips.
The scheme you propose (I am guessing what :X means) could be a problem for
further enhancements if it captures _more_ than the erroneous addresses.
An error-specification scheme that is close to hardware [like in my scheme
of address selection] makes it possible to narrow the discarded memory to
fit snugly around the error bits, for instance to provide slabs.
May I ask: What is the intention of this "smap" option; is it to use
damaged RAM or does it serve another purpose?
Cheers,
-Rick.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Mar 23 2000 - 21:00:31 EST