On Thu, 21 Aug 1997, Brad Spencer wrote:

   > Given a Sparc 1 with a external hard drive and a Conner tape changer
   > [DDS-2 drive with 4 tapes, using 90M tapes].  Every so often, for reasons
   > which do not make complete sense, I get the following SCSI errors:
   > 
   > esp0: RESELECT: 0 bytes in FIFO! [intr c, stat 17, step 4, prevphase 3]

   Looks like the famous esp reselect bug.  It's a deficiency in the hardware
   which pops up when a reselecting device interrupts the esp driver during a
   command loading phase.  You'll notice it happens whenever there is a
   moderate amount of activity to more than one SCSI target, of which one has
   re-selection enabled.

I see, how annoying.  This would be the case, the hard disk has reselect
disabled [the default for id 0 to 3], the tape changer has it enabled [as
is the default for ids 4 to 6].  I'll move it to a id which has reselects
disabled and see what happens.

   There has been an attempted workaround to this problem in -current, but I
   haven't heard from anyone who's tried it.

On the machine in question, I am running a fairly recent -current, within
a day or two.  If the workaround is on automatically, then there appears
to be more problems.

   -J

Thanks....


Brad Spencer - brad@anduin.eldar.org   http://anduin.eldar.org