On Wed, 5 Apr 2000, Mark Hahn wrote:
> > None of the PC stuff ever corrected RAM errors with a CRC, even though
> nonsense. the BX chipset, for instance, most definitely supports SECDED;
> see section 3.3.28 in the datasheet. I believe the LX did as well,
> and the HX back in the P5 era. RDRAM seems to be _only_ available in ECC.
bx and lx are very similar in terms of ecc support. bx being slightly
better in that it has a specific register to report the location of errors
down to a granularity of 4k blocks.
> > ECC RAM just accumulated bits from multiple columns to save parity
> > bits. It still hit MNI when it detected an error. Alpha and Suns
> BX is normally configured (if ECCing) to report only uncorrected errors.
> there's a chipset register to report the problem address, and you can
> tell it to report even corrected, single-bit errors. AFAIK none of this
> is touched by Linux.
Not only that; the bx can auto-correct errors, no memory scrubbing
required.
Annoying: 440zx is identical to 440bx except it doesnt support ecc.
However there seems to be no way to detect 440bx from 440zx in software!
:-( Same problem with 440lx (ecc) and 440ex (non ecc). :-(
-Dan
-
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 : Fri Apr 07 2000 - 21:00:15 EST