5.3 on ultra2: scsi disk not detected

[Scott Long]

mk at capri.pl wrote:
> [...]
> 
> Linux infinitely keeps repeating last warning. FreeBSD at least reports 
> few parity errors, skips the disk and goes on. 
> 
> For now I'm out of spare SCA disks, so tests stop here, but I will try 
> some more disks when I find them, because it's intriguing.
> 
> FreeBSD/Ultra2 users: what disks work for you ? Are the disks exactly 
> what Sun describes as "compatible", or maybe something else works also ?
> Not for me :( I thought SCSI is SCSI, especially considering the same 
> generation disks.
> 
> Best regards,
> 

u2# camcontrol devlist -v
scbus0 on esp0 bus 0:
<SEAGATE ST34371W SUN4.2G 7462>    at scbus0 target 0 lun 0 (da0,pass0)
<IBM DNES30917SUN9.0G SAD0>        at scbus0 target 1 lun 0 (da1,pass1)
<TOSHIBA CD-ROM XM-3701TA 0236>    at scbus0 target 6 lun 0 (cd0,pass2)
<  >                               at scbus0 target -1 lun -1 ()
scbus-1 on xpt0 bus 0:
<  >                               at scbus-1 target -1 lun -1 (xpt0)


I boot FreeBSD off of da0 and NetBSD off of da1.  Both can see all three
devices.  So it's not a problem that is related to the number of disks.
My guess here, since I'm not a Solaris expert and don't know how to
extract the information out of it that would be helpful, is that either
your misbehaving disk is 'special' and solaris knows to do special
things with it, or it's stuck in narrow mode or has a faulty pin and
solaris is smart enough to deal with that.  Neither FreeBSD nor Linux
know how to do domain validation, so a faulty pin on the upper half of
the bus would show up as a mystery parity error.  A faulty pin on the
lower half of the bus could interfer with selections on the other 
low-numbered devices, though I have no idea how solaris could compensate
for this.

I can hack the driver to allow you to force everything into async-narrow
mode.  That might help identify the problem, but it won't help much if
it's a case of the drive needing special instructions that we don't
understand.  A SCSI bus capture would be ideal.

The only other idea is that the disk is auto-terminating the bus and
interfering with the termination that already exists.  Again, I have no
idea how solaris could compensate for this unless it's forcing 
everything to async, or it knows how to 'fix' the misbehaving device.

Scott