I went through the ahd code looking for a clue and the only thing that would make a difference is if the chip/bus hasn't settled yet. If I read the code right the only way that can happen is when it hits a SCSI reset on channel A. This really shouldn't happen so that makes your hardware suspect. Can you send a dmesg of the offending system please?
On Sun, Aug 17, 2008 at 03:52:10AM +0200, Christoph Viethen wrote: > Hello all, > > had a little problem getting OpenBSD to run on one of our machines. > > It has a SCSI adapter nicely supported by the ahd(4) driver, but the > problem I had was that all SCSI devices would show up, with the > exception of - ta dah! - the hard disk. The only one in the machine, > supposed to be the boot disk. Now that didn't make installation too > easy, but applying some creativity (and, temporarily, an extra ATA > drive), it could be done. :-) > > The solution was to set the SCSI_DELAY option of the kernel to some > larger value (tried 15 seconds, and that worked fine - didn't have the > nerve to figure out what value would have been the minimal one which > still would solve the problem). This way, the whole SCSI initialization > process would pause for a while, allowing the harddisk to turn up. > > Now, the whole issue obviously is not really OpenBSD's fault - when a > harddisk is too slow to turn up, well, bad luck. What I'm wondering, in > any case, is, whether my approach to solving this problem was the right > one, or whether there would have been an easier way. I mean, I would have > preferred to set some options in UKC or with config(8), but I was > clueless as to whether setting the SCSI delay can be achieved this way, > or whether building a special kernel was the only approach. > > Could somebody please shed some light on this? Is building a new kernel > the only thing one can do when one suffers from a too-slow SCSI device, > or might there be a shortcut (other than throwing the SCSI adapter and/or > drive out of the window and getting a different one, of course) ? > > > Thanks, > > Christoph > > -- > [EMAIL PROTECTED]
