On Sun, Nov 02, 2003 at 02:17:29PM +0100, Manuel Bouyer wrote:
> On Fri, Oct 31, 2003 at 04:32:51PM +0200, Jukka Marin wrote:
> > The command completes OK, but if I fsck the CF card, I get:
> > 
> > # fsck -f /dev/rwd1a
> > ** /dev/rwd1a
> > BAD SUPER BLOCK: VALUES IN SUPER BLOCK DISAGREE WITH THOSE IN FIRST ALTERNATE
> > 
> > LOOK FOR ALTERNATE SUPERBLOCKS? [yn] n

After sending this message, I found out that reading the CF works OK, but
writing is unreliable.

> > The same card and same image works OK on an HP laptop.  A different CF
> > card works OK on the HP _and_ my laptop.  Both cards are Kingston 128 MB
> > cards.  The ones that do now work are a bit newer:

Should read ".. do NOT work.." ;)

> > wd1 at atabus2 drive 0: <SAMSUNG CF/ATA>
> > wd1: drive supports 4-sector PIO transfers, LBA addressing
> > wd1: 124 MB, 496 cyl, 16 head, 32 sec, 512 bytes/sect x 253952 sectors
> > 
> > The older cards look like this:
> > 
> > wd1 at atabus2 drive 0: <TOSHIBA THNCF128MMA>
> > wd1: drive supports 1-sector PIO transfers, LBA addressing
> > wd1: 122 MB, 978 cyl, 8 head, 32 sec, 512 bytes/sect x 250368 sectors
> > 
> > The problem existed with an older kernel (1.6W) as well.
> 
> Could you try to force it to 1-sector PIO transfer ?
> Just force wd->sc_multi to 1 in wd.c:wdattach() line 326

This fixes the problem.  So I guess the Samsung cards are lying when they
claim to support 4-sector transfers?  Or is there a bug in the NetBSD driver
or something?

Why doesn't this problem show up on a slower HP laptop, but only on my
2 GHz IBM?

  -jm

P.S. Using 1-sector transfers on the real HD is S L O W ;-)