--Signature_Thu__9_Jun_2005_07_25_36_-0400_sg.Dn2kvdTBUSlrB
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable

Hello,

> > It would be good if NetBSD could use, store, and maintain a "bad page=20
> > map" so that once a bad page in RAM is identified (through parity/ECC=20
> > errors), it can be avoided.
>=20
> Hmm, if a parity error happens, does it mean it's going to happen again a=
nd
> again ?

Most likely. And if ECC can't cover it up you'll get all kinds of nastiness=
 from it.=20

> How can we be sure that reading/writing again to the same RAM location
> will yield the same error ?

We can't. If it's just one bit that's stuck you won't even get a parity err=
or every time you access it ( because it would be in the right position hal=
f the time )

But adding code that checks for a stuck bit in the NMI handler and then dis=
ables the page ( or puts it on a bad list or whatever ) wouldn't be too har=
d I guess.

have fun
Michael

--Signature_Thu__9_Jun_2005_07_25_36_-0400_sg.Dn2kvdTBUSlrB
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (NetBSD)

iQEVAwUBQqgnMcpnzkX8Yg2nAQJ+IQf/Vhb6KiX35lBDzHXhfzsMqstf7ag64JDb
hjfByrPujPVUWMF1mChWCa2uvdbYS9fVNECClabmqbxcjz76m+376aCfW6ZLwrFs
aPw9hEfveTc6H8e5j7QAvAKB+SKDBSfGB/gRoJqKL6c2ham34R9AqzAuCFGnPO5q
OHrQufcsRHVpu1njdD8kZNhdWhbtWjwrVh21MWXci0MQPwbHV8x8Dcv2w3oCcnIC
ToFuogWjBEmfLeDW0c6ghE5/v/mOCGkFzTi7rm6OrZ7xsy+mAUnvrbgQrgpadSkP
63BReOoA0pIS7TCn5PVI4Ymtgl3P01tRsQNkJH2o88CJhhfghaNwYg==
=Smh4
-----END PGP SIGNATURE-----

--Signature_Thu__9_Jun_2005_07_25_36_-0400_sg.Dn2kvdTBUSlrB--