tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]


On Mon, Dec 07, 2009 at 03:23:11AM -0500, der Mouse wrote:
> >> I would not suggest doing this unless (a) the drive claims to be
> >> LBA48-capable and (b) the drive claims to be at least 0x10000000
> >> sectors long.
> > Make that 0x10000001.  128Gb drives don't need LBA48 to access the
> > last sector.
> In theory, maybe.  As I remarked upthread, I own a drive which errors
> on non-LBA48 accesses to sector 0xfffffff. (Admittedly, it's
> substantially over 128G, but I see no reason a drive of exactly 128G
> couldn't exhibit the same syndrome.)

Yes, I'm aware such drives exists.

> > But then you'd break actually-working setups.
> The only setups such a change would break would be those with >128G
> drives onj non-LBA48 controllers but which do not actually access the
> drive beyond the 128G mark.
> Even that much could be avoided by the "cap at 128G and fall back to
> non-LBA48 on error", unless there exist interfaces which react to LBA48
> accesses with neither correct LBA48 operation nor an easily detectible
> and retryable error.  Er, unless such interfaces exist and are common
> enough to care about - we can't work around _all_ broken hardware.

It's not broken. These controllers were designed before LBA48
was specified.

> The "try it and see if it errors" approach could maybe also be used to
> detect the "broken 0xfffffff" syndrome too....

This is what is actually done

Manuel Bouyer <>
     NetBSD: 26 ans d'experience feront toujours la difference

Home | Main Index | Thread Index | Old Index